A method, equipment, and medium for assessing typhoon disaster and disaster resistance capabilities of offshore wind turbines based on artificial intelligence.

Abstract

This application discloses an artificial intelligence-based method, equipment, and medium for assessing typhoon disaster and disaster resistance capabilities of offshore wind turbines, relating to the field of intelligent assessment. The method includes: extracting damage features from multimodal data of the offshore wind turbine and generating structured damage information; performing intent recognition, parameter extraction, and parameter completion based on user commands and a large language model to obtain a standardized disaster analysis simulation parameter set; calling an integrated wind power load simulation tool to simulate the structural dynamic response under typhoon conditions, and performing closed-loop correction based on on-site measured data; and generating an assessment report based on damage identification results, simulation analysis results, and disaster resistance capability evaluation indicators. The disaster resistance capability evaluation indicators include at least one of the following: maximum equivalent stress at the tower base, stress safety margin, recommended pitch angle, or stress reduction rate.

Description

Technical Field

[0001] This application relates to the field of intelligent assessment, and in particular to a method, equipment, and medium for assessing the typhoon disaster and disaster resistance capabilities of offshore wind turbines based on artificial intelligence. Background Technology

[0002] With the transformation of the global energy structure, wind energy, as a clean and renewable energy source, is receiving increasing attention for its development and utilization. As a key piece of equipment for wind energy conversion, the safety and stability of wind turbines directly affect the operational efficiency and economic benefits of wind farms. However, under extreme weather conditions such as typhoons, wind turbines are susceptible to severe damage, leading to decreased power generation efficiency or even shutdown. Therefore, disaster assessment of wind turbines and timely prediction and repair of damage are of great significance for ensuring the stable operation of wind farms.

[0003] Currently, most wind turbine disaster assessment solutions on the market rely on multiple independent tools, each responsible for different tasks, such as image detection, wind turbine damage analysis, and wind farm simulation. While these tools are effective in their respective fields, they suffer from the following major problems: 1. Lack of integration: Existing assessment solutions often use different tools in a fragmented manner, requiring multiple professionals to operate different systems. This results in poor data flow, difficulties in collaboration between different tools, and often delayed feedback from field data, making real-time updates and corrections impossible, thus reducing the efficiency and accuracy of the assessment.

[0004] 2. Inability to Achieve Intelligent Assessment (i.e., Low Level of Intelligence): Traditional wind turbine disaster assessment relies on high-precision simulation tools. While these tools can perform detailed simulation calculations, they often require complex data input and highly specialized operation. Furthermore, these tools typically lack intelligent capabilities, failing to automate task configuration and real-time data analysis. Especially under extreme weather conditions such as typhoons, the assessment process requires complex data input and highly specialized operation, resulting in low assessment efficiency and accuracy.

[0005] 3. Insufficient real-time capability: Existing assessment schemes cannot compare simulation results with field data in real time, and lack a feedback mechanism to automatically optimize the assessment results. Under rapidly changing weather conditions such as typhoons, the assessment results may not reflect the actual situation in a timely manner, leading to a lag in the formulation of remediation strategies.

[0006] Based on the above description, while existing tools are effective in their respective fields, they lack integration, requiring multiple professionals to operate different systems. This results in poor data flow, difficulties in collaboration, and often delayed feedback from on-site data, hindering real-time updates and corrections. Furthermore, traditional assessment methods rely on high-precision simulation tools, but these tools typically lack intelligent capabilities, failing to automate task configuration and real-time data analysis, leading to low efficiency and accuracy, especially under extreme weather conditions such as typhoons. Moreover, current technologies often struggle to simultaneously achieve damage detection, natural language task parsing, load simulation invocation, and closed-loop feedback of on-site data within the same workflow, completing all tasks on a single integrated platform, particularly for wind turbine disaster assessment under extreme weather conditions like typhoons. Existing solutions require multiple tools to execute different task steps, resulting in low overall workflow efficiency, poor accuracy, slow response, and a lack of real-time correction and feedback of on-site data, failing to provide accurate and timely disaster assessments. Therefore, how to achieve an integrated setup to complete the entire process from typhoon disaster assessment to wind turbine status analysis, thereby addressing these problems in existing technologies, has become a pressing technical challenge in this field. Summary of the Invention

[0007] The purpose of this application is to provide an artificial intelligence-based method, equipment, and medium for assessing the typhoon disaster and disaster resistance capabilities of offshore wind turbines. This method can correct and provide feedback on on-site data in real time, providing accurate and timely disaster assessments, thereby solving the problems of low overall workflow efficiency, poor accuracy, and slow response in existing technologies.

[0008] To achieve the above objectives, this application provides the following solution: Firstly, this application provides a method for assessing the typhoon disaster and disaster resistance capabilities of offshore wind turbines, including: Based on multimodal data from offshore wind turbines, damage features are extracted using computer vision and target detection algorithms to generate structured damage information. The structured damage information includes at least one of the following: damaged component identification, damage category, damage location, damage scale, severity level, detection confidence, data source, acquisition time, and associated image number. Based on user commands, the semantic understanding capabilities of a large language model are combined to perform intent recognition and key parameter extraction, resulting in a classification task and a standardized disaster analysis simulation parameter set. The standardized disaster analysis simulation parameter set includes explicit parameters in the user commands, damage state parameters determined by structured damage information, and environmental condition parameters supplemented through interactive questioning or external data interfaces. User commands are categorized into tasks; when the categorized task involves disaster simulation analysis, a standardized disaster analysis simulation parameter set is generated based on user commands, structured damage information, and environmental condition data, and the integrated wind power load simulation tool is called to perform the simulation analysis. Based on the answer or simulation analysis result corresponding to the user command, an assessment report on typhoon disaster and disaster resistance capability of offshore wind turbines is generated; wherein, the assessment report on typhoon disaster and disaster resistance capability of offshore wind turbines includes at least one of the following: damage identification results, structural dynamic response results, on-site measured data comparison results, disaster resistance capability evaluation indicators, and typhoon operating condition control strategy recommendations.

[0009] Secondly, this application provides an equipment for assessing the typhoon disaster and disaster resistance capabilities of offshore wind turbines, including: A damage detector is used to extract damage features and generate structured damage information based on multimodal data of offshore wind turbines using computer vision and target detection algorithms. The structured damage information includes at least one of the following: damaged component identification, damage category, damage location, damage scale, severity level, detection confidence, data source, acquisition time, and associated image number. The user interactor is used to perform intent recognition and key parameter extraction based on user commands and the semantic understanding capabilities of a large language model, to obtain a classification task and a standardized disaster analysis simulation parameter set. The standardized disaster analysis simulation parameter set includes explicit parameters in user commands, damage state parameters determined by structured damage information, and environmental condition parameters supplemented by interactive questioning or external data interfaces. Information analysis and task identifier, used to classify user commands into tasks; The wind load simulator is used to generate a standardized disaster analysis simulation parameter set based on user instructions, structured damage information and environmental condition data when the classification task involves disaster simulation analysis, and then call the integrated wind load simulation tool to perform the simulation analysis. The result feedback device is used to form a closed-loop feedback based on the simulation analysis results, and to trigger simulation parameter correction and re-simulation when the difference between the simulation results and the field measured data does not meet the preset consistency criteria; the result feedback device is also used to generate an assessment report on typhoon disaster and disaster resistance of offshore wind turbines based on at least one of the answers corresponding to the user instructions, simulation analysis results, disaster resistance evaluation indicators and typhoon operating condition control strategy suggestions. The damage detector, the user interactor, the information analysis and task identifier, the wind power load simulator, and the result feedback device are all integrated and deployed on the same platform, and run in series with the intelligent agent workflow through a unified software architecture.

[0010] Thirdly, this application provides a computer device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the above-mentioned method for assessing the typhoon disaster and disaster resistance capabilities of offshore wind turbines.

[0011] Fourthly, this application provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the steps of the above-described method for assessing the typhoon disaster and disaster resistance capabilities of offshore wind turbines.

[0012] According to the specific embodiments provided in this application, this application has the following technical effects: This application provides an artificial intelligence-based method, equipment, and medium for assessing the typhoon disaster and disaster resistance capabilities of offshore wind turbines. The solution integrates damage detection, natural language interaction, task scheduling, load simulation, and closed-loop feedback into a single software architecture, reducing the need for switching between multiple tools and manual parameter processing, and improving the automation level of the disaster assessment process. The solution utilizes a large language model to parse natural language or voice commands, recognize intent, and complete parameters, enabling users to initiate disaster assessment tasks using natural language. Furthermore, the solution triggers simulation parameter correction through preset consistency criteria, improving the consistency between simulation results and field observation data. Further, based on disaster resistance evaluation indicators such as maximum equivalent stress at the tower base, structural stress margin, recommended pitch angle, and stress reduction rate, the solution provides data support for assessing the disaster resistance capabilities and making control strategy decisions for offshore wind turbines under typhoon disaster conditions. Attached Figure Description

[0013] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0014] Figure 1 A flowchart illustrating a method for assessing typhoon disasters and disaster resistance capabilities of offshore wind turbines, provided as an embodiment of this application; Figure 2 A schematic diagram illustrating the process of wind turbine damage detection using a YOLO model and a large visual model, as provided in an embodiment of this application; Figure 3 A schematic diagram illustrating the implementation process of an offshore wind turbine typhoon disaster and disaster resistance capability assessment device provided in an embodiment of this application; Figure 4 A flowchart illustrating the process of extracting disaster analysis simulation parameters from user interaction, provided as an embodiment of this application; Figure 5 A schematic diagram of an integrated analysis theory and design analysis software tool provided in an embodiment of this application; Figure 6 This is a schematic diagram of the structure of a computer device provided in an embodiment of this application. Detailed Implementation

[0015] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[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] In one exemplary embodiment, this application provides a method for assessing the typhoon disaster and disaster resistance capabilities of offshore wind turbines. This method is executed by computer equipment, specifically by a terminal or server, or by both. In this embodiment, the method is described using a server as an example. Figure 1 As shown, the method includes: Step 100: Based on the multimodal data of offshore wind turbines, damage features are extracted using computer vision and target detection algorithms to generate structured damage information. The structured damage information includes at least one of the following: damaged component identification, damage category, damage location, damage scale, severity level, detection confidence level, data source, acquisition time, and associated image number.

[0018] Step 101: Based on user commands, and leveraging the semantic understanding capabilities of a large language model, perform intent recognition and key parameter extraction to obtain a classification task and a standardized disaster analysis simulation parameter set. The standardized disaster analysis simulation parameter set includes explicit parameters from user commands, damage state parameters determined by structured damage information, and environmental condition parameters supplemented through interactive questioning or external data interfaces. Specifically, the damage state parameters determined by structured damage information may include at least one of the following: damaged component, damage location, damage category, damage scale, damage level, and stiffness reduction factor; the environmental condition parameters supplemented through interactive questioning or external data interfaces may include at least one of the following: typhoon path, wind speed, wind direction, wave parameters, tidal current parameters, unit model, and control status.

[0019] Step 102: Classify user instructions into tasks; when the classified task involves disaster simulation analysis, generate a standardized disaster analysis simulation parameter set based on user instructions, structured damage information and environmental condition data, call the integrated wind power load simulation tool to perform structural dynamic response simulation under typhoon conditions, and compare the simulation results with the field measured data; when the difference between the simulation results and the field measured data does not meet the preset consistency criteria, correct the simulation parameters and re-perform the simulation.

[0020] Step 103: When the difference between the simulation results and the field measured data meets the preset consistency criteria, calculate the disaster resistance evaluation index based on the simulation analysis results, and generate an assessment report on typhoon disaster and disaster resistance capabilities of offshore wind turbines. The disaster resistance evaluation index includes at least one of the following: maximum equivalent stress at the tower base, stress safety margin, recommended pitch angle under typhoon conditions, or stress reduction rate relative to the benchmark control strategy; the assessment report includes at least one of the following: damage identification results, structural dynamic response results, comparison results of field measured data, disaster resistance evaluation index, and typhoon condition control strategy recommendations.

[0021] In one exemplary embodiment of this application, the implementation process of step 100 above in order to achieve accurate damage detection includes: Step 100-1: Using the YOLO model, target localization and initial damage category identification are performed on the multimodal data to obtain the initial damage area and category.

[0022] Step 100-2: Using a large visual model, based on the damage area and category, assess the severity of the damage, generate semantic interpretation and treatment suggestions, and obtain structured damage information.

[0023] In one exemplary embodiment of this application, in order to further improve the accuracy of damage detection, it is necessary to train a YOLO model, the process of which may include: Step 1: Data collection and labeling.

[0024] A large amount of image data of wind turbines needs to be collected, including different types of damage such as burns, cracks, deformation, and peeling. This image data mainly comes from on-site shooting and publicly available datasets.

[0025] The collected image data is labeled to identify the damaged areas and types in the wind turbine image data, forming the basis of the training set. Damage labels include categories such as cracks, burns, and deformation, providing necessary annotation information for the YOLO model's training dataset.

[0026] Step 2: Dataset partitioning.

[0027] All collected damage image data were divided into training, validation, and test sets in a 7:2:1 ratio. This data partitioning strategy ensures the efficiency of the training process and the model's generalization ability. The training set is used for model training, iteratively optimizing the model's parameters. The validation set is used for parameter tuning during training to ensure the model does not overfit. The test set is used for the final evaluation of the model after training, verifying its generalization ability and accuracy.

[0028] Step 3: YOLO model training.

[0029] Target detection is performed using the YOLO model. The YOLO model extracts features from images, quickly identifying and locating target regions. It employs a regression problem to predict the location and category of targets in an image, enabling efficient and accurate detection of damage targets in wind turbines at different scales.

[0030] After multiple rounds of training and optimization, the YOLO model successfully completed the wind turbine damage detection task, which can quickly identify and extract damage information of wind turbines and improve the accuracy and efficiency of detection.

[0031] Step 4: Call the tool via the Model Context Protocol (MCP).

[0032] The trained YOLO model can be invoked through MCP to work collaboratively with other modules within the platform. Users can trigger the model through natural language input or image uploads, and the YOLO model will then be invoked to perform damage detection based on the input data.

[0033] Based on the above description, in practical applications, the entire damage detection process can be described as follows: (1) Data upload and processing: Users upload on-site images of wind turbines or relevant meteorological data to the platform, or perform damage detection directly through voice commands.

[0034] (2) Image preprocessing and detection: The uploaded image data is preprocessed using the YOLO model to remove noise and improve image quality before damage detection. The YOLO model quickly locates possible damage areas in the image, such as cracks and deformations, using the damage label model obtained through training.

[0035] (3) Damage detection and feedback: The YOLO model outputs the detection results in real time, and further analyzes and classifies the damage types based on the damage information, providing accurate information for subsequent simulation and disaster assessment.

[0036] (4) Results Display and User Interaction: The detected damage information is displayed to the user in a visual manner, marking the damaged areas and providing specific damage types and locations. Through interaction with the user, the user can provide feedback on the results and then optimize and iterate.

[0037] Based on the above description, the two-stage wind turbine damage detection process of YOLO model + visual large model provided in this application is as follows: Figure 2 As shown. Figure 2 The curves and values ​​shown are for illustrative purposes only and do not constitute a limitation on the scope of protection of this application.

[0038] In one exemplary embodiment of this application, in addition to initiating tasks by inputting text, uploading images, or weather data via keyboard, users can also issue control commands to the platform via voice. This voice control process is not simply a matter of recording and executing directly; rather, it involves multiple stages such as voice processing, text conversion, semantic understanding, task recognition, and tool scheduling before finally forming executable system commands. Its core can be divided into the following two parts: I. Clearly identify user input through voice processing.

[0039] In this implementation, voice capability is achieved by calling the speech recognition model interface to convert processed voice commands into text. Voice processing can be understood as first converting sound into standardized text, and then handing the text over to the platform for analysis, rather than having a human interpret what the user said. The specific process is as follows: 1. Voice capture.

[0040] Users input voice commands via microphone on mobile phones, tablets, computers, or other smart terminals, such as "Please perform damage detection on this wind turbine photo," "Perform load simulation based on the current typhoon path," and "Extract blade crack information and generate an evaluation task." The system collects the user's raw voice signal and sends it to the voice processing link.

[0041] 2. Speech preprocessing.

[0042] Because real-world environments may contain wind noise, equipment background noise, and unstable speaking volume, raw speech cannot usually be used directly for recognition. Therefore, the speech signal is first preprocessed, including: 1) Amplification processing: Enhance weak speech to make the main body of the speech clearer.

[0043] 2) Noise reduction: Reduce interference from environmental noise, wind noise, and current noise.

[0044] 3) Necessary audio integration processing: Organize the collected audio segments to make them more suitable for subsequent speech recognition.

[0045] The goal of this stage is not to understand the meaning of the instructions, but to improve the quality of the speech signal and reduce subsequent recognition errors.

[0046] 3. Speech-to-text conversion.

[0047] After preprocessing, the speech content is converted into standardized text using the speech recognition capabilities provided by the speech recognition model interface. This is essentially the Automatic Speech Recognition (ASR) process. After this process, the original audio signal is transcribed into text content for further processing, such as converting the user's speech into text like "perform damage detection on the uploaded image," "extract wind speed and typhoon path parameters and perform simulation," and "generate disaster analysis tasks by combining on-site photos and meteorological data."

[0048] 4. Text standardization.

[0049] Even after speech recognition, the resulting text may still contain issues such as colloquial expressions, pauses, omissions, and non-standard phrasing. Therefore, the recognition results can be standardized. For example, colloquial expressions like "Help me check if the blades of this wind turbine are cracked" can be transformed into more standardized task descriptions, such as "Identify whether there are cracks or damage on the blades of a wind turbine." The purpose of this is to make subsequent large-scale model analyses more stable and accurate.

[0050] Therefore, from a technical implementation perspective, voice input does not directly enter the simulation or detection. Instead, it first calls the cloud-based speech recognition service, and after voice acquisition, signal preprocessing, ASR speech recognition, and text standardization, it is converted into standardized text commands before entering the workflow of this application.

[0051] 2. Identify the corresponding instructions.

[0052] After the speech has been converted into standardized text, the task is not determined by fixed buttons or manual selection, but by a large model acting as the "central analysis brain," performing semantic understanding, intent recognition, parameter extraction, and task distribution on the text content. This processing logic is consistent with the "natural language understanding module" of the user interface and the "task classification and scheduling module" of the Recognizer. Based on this, it can be divided into the following steps: 1. Semantic understanding.

[0053] The large language model service interface is invoked to perform semantic analysis on the text output by ASR. Here, it identifies not individual words, but the task intent expressed by the entire sentence. For example, a user might say, "Help me identify if there is any obvious damage to the tower in this photo," "Consider the typhoon's path and calculate the stress on this wind turbine," or "Identify cracks first, then perform a disaster analysis." The large model will understand that these statements correspond to "damage detection tasks," "simulation analysis tasks," or "combined multi-step tasks," rather than simply identifying a few isolated keywords.

[0054] 2. User intent recognition.

[0055] Building upon semantic understanding, the large model further determines what the user truly wants to do, that is, identifies the task intent. Common intents include at least: initiating damage detection; extracting disaster analysis parameters; initiating wind power load simulation; querying analysis results; and triggering the next task based on the results.

[0056] 3. Key parameter extraction.

[0057] Large models also extract key parameters from text. For example, they can identify from user commands: wind turbine type, typhoon path, wind speed and direction, specified images or field data, damage locations, simulation range, or analysis targets. These key parameters are then organized into structured, callable information for direct use later.

[0058] 4. Task classification and scheduling.

[0059] In this application, tasks are classified based on the output of a large model. For example, if a user says "What damage is shown in this image?", it will be identified as a damage detection / analysis-oriented task. If a user says "Simulate the load response of this wind turbine based on typhoon parameters", it will be identified as a simulation task. If a user says "Identify the damage first, then provide a disaster assessment", it will be identified as a combined task.

[0060] 5. The output is an executable instruction.

[0061] The final output of the large model is not plain text for users, but rather task descriptions or invocation instructions that can be executed within the platform. In other words, although users make requests through natural language or even spoken language, once inside the platform, they are transformed into standardized, structured, and schedulable task commands, which are then handed over to the corresponding modules for execution.

[0062] Therefore, the so-called identification of the corresponding instruction is not actually accomplished by simple keyword matching. Instead, it is achieved by using a large model to perform semantic understanding, intent judgment, and parameter extraction on the user's statement, and then using task identification and scheduling to map it into a specific function call path.

[0063] In an exemplary embodiment of this application, step 101 above, which involves extracting key parameters based on user instructions and combining the semantic understanding capabilities of a large language model to obtain a standardized disaster analysis simulation parameter set, includes: Step 101-1: Extract explicit parameters from user instructions based on the semantic understanding capability of the large language model, and detect missing key parameters in the simulation based on structured damage information.

[0064] Step 101-2: Complete the missing key parameters by interactively probing or calling external data interfaces to obtain the completed parameters. Step 101-3: Standardize and standardize the explicit parameters, the damage state parameters determined by structured damage information, and the completed parameters to generate a standardized disaster analysis simulation parameter set.

[0065] In one exemplary embodiment of this application, the integrated wind power load simulation tool employs an integrated coupled analysis theory of aerodynamics-hydraulics-soil dynamics-multibody-servo control for simulation. For example, the simulation process includes: generating aerodynamic and hydrodynamic loads based on meteorological and marine environmental data; establishing pile-soil coupling boundary conditions; solving the overall dynamic response from the wind turbine to the foundation; and embedding an adaptive control strategy module in the solution chain to dynamically adjust the unit's operating state according to the real-time load response, achieving bidirectional coupled calculation of control and structural response.

[0066] In one exemplary embodiment of this application, the implementation process of step 102 described above may include: Step 102-1: When the classification task is a reasoning task, the reasoning agent is invoked to generate the answer corresponding to the user instruction; Step 102-2: When the classification task is a tool invocation task, the integrated wind power load simulation tool is invoked to simulate the structural dynamic response of offshore wind turbines under typhoon conditions and evaluate control strategies based on a standardized disaster analysis simulation parameter set. When the difference between the simulation results and the field measured data meets the preset consistency criterion, the simulation analysis results are obtained; wherein: Steps 102-11: Simulate and evaluate the structural dynamic response and control strategy of offshore wind turbines under typhoon conditions based on the standardized disaster analysis simulation parameter set, and obtain simulation results.

[0067] Steps 102-12: Compare and analyze the simulation results with the field measured data to obtain the comparison results. The field measured data includes at least one of the following: unit operation monitoring data, structural response monitoring data, environmental condition data, and inspection image data, such as SCADA data, strain sensor data, acceleration sensor data, tower tilt angle data, UAV inspection images, on-site photos, wind speed and direction data, typhoon path data, wave data, tidal current data, or water level data, etc.

[0068] Steps 102-13: Determine whether the comparison results exceed the preset thresholds and obtain the judgment result. When the judgment result is yes, trigger the closed-loop feedback mechanism, correct the simulation parameters, and re-perform the simulation. When the judgment result is no, obtain the simulation analysis result. For example, the comparison results not exceeding the preset thresholds can be described as: the tower top displacement error does not exceed the preset threshold; the blade root load error does not exceed the preset threshold; the tower bottom bending moment error does not exceed the preset threshold; the damage location matching degree reaches the preset threshold; the damage level difference does not exceed the preset level range, etc.

[0069] In some embodiments, after obtaining simulation analysis results, the equipment calculates disaster resilience evaluation indicators based on the simulation analysis results. The disaster resilience evaluation indicators include at least one of the following: maximum equivalent stress at the tower base, structural stress margin, damage severity level, recommended control strategy, and stress reduction rate. Specifically, the maximum equivalent stress at the tower base characterizes the structural response strength of the offshore wind turbine tower base under typhoon conditions; the stress safety margin or structural stress margin characterizes the margin of the offshore wind turbine structural response from a preset allowable stress threshold under typhoon conditions, and the stress safety margin or structural stress margin is determined based on the difference or ratio between the preset allowable stress threshold and the maximum equivalent stress at the tower base; the recommended control strategy may include a recommended pitch angle corresponding to the yaw angle under typhoon conditions; and the stress reduction rate can be determined based on the difference between the maximum equivalent stress at the tower base under the baseline control strategy and the maximum equivalent stress at the tower base under the recommended control strategy. Thus, the equipment can transform damage detection results, load simulation results, and field data comparison results into disaster resilience evaluation results that can be used for typhoon disaster decision-making.

[0070] Based on the same inventive concept, this application also provides an offshore wind turbine typhoon disaster and disaster resistance assessment device for implementing the aforementioned method for assessing the typhoon disaster and disaster resistance capabilities of offshore wind turbines. The solution provided by this device is similar to the solution described in the above method. Therefore, the specific limitations of one or more embodiments of the offshore wind turbine typhoon disaster and disaster resistance assessment device provided below can be found in the limitations of the offshore wind turbine typhoon disaster and disaster resistance capability assessment method described above, and will not be repeated here.

[0071] In one exemplary embodiment, such as Figure 3 As shown, the provided equipment for assessing the typhoon disaster and disaster resistance capabilities of offshore wind turbines includes: a damage detector, an interrogator, an information analyzer and task recognizer, a wind load simulator, and a result feedback device. Each component is modularly processed and integrated into a single platform, operating in series with the intelligent agent workflow through a unified software architecture.

[0072] Based on this, the functional descriptions of each module are as follows: (1) Damage detector.

[0073] The damage detector is configured to extract damage features and generate structured damage information based on multimodal data from offshore wind turbines using computer vision and target detection algorithms. For example, the core function of this module is to extract damage evidence of wind turbines from multi-source data such as on-site images, meteorological data, and satellite images, including crack height, collapse direction, and fault type.

[0074] (2) User Interactor.

[0075] The user interface is configured to perform intent recognition and key parameter extraction based on user commands, combined with the semantic understanding capabilities of a large language model (hereinafter referred to as the large model), to obtain a classification task and a standardized disaster analysis simulation parameter set. The standardized disaster analysis simulation parameter set includes explicit parameters in user commands, damage state parameters determined by structured damage information, and environmental condition parameters supplemented through interactive questioning or external data interfaces.

[0076] like Figure 4 As shown, this module provides user interaction, receiving user natural language commands and performing preliminary intent recognition and extraction of disaster analysis simulation parameters. Users submit their typhoon disaster assessment requirements for wind turbines via natural language, such as specifying the typhoon path, wind speed, and turbine type. The implementation process includes: A. User input and natural language understanding.

[0077] After damage detection is completed, the damaged area, type, and location are displayed to the user, serving as the basis for subsequent interactions. The natural language input received by the user interrogator typically arises after the user has seen the aforementioned on-site damage information. That is, the user first sees the identified damage results and then wants to understand the type, severity, possible causes, and whether the differences between the damage and the typhoon disaster simulation results meet the preset consistency criteria. Therefore, the user interacts with the platform through natural language. At this point, the user interrogator uses its natural language understanding module to perform semantic analysis on the user's statements, identifying whether the user is querying damage information, requesting explanations from the platform, or hoping to further conduct typhoon disaster simulation analysis.

[0078] If the user simply wants to understand the on-site damage information, the device will recognize the input as an explanation or analysis task and provide explanations to the user based on existing damage detection results. If the user wants to further utilize simulation to confirm whether the difference between the on-site damage and the typhoon simulation results meets preset consistency criteria, the device will recognize the input as a simulation-related task and guide the user to supplement the necessary data required for the simulation, such as typhoon path, wind speed, wind direction, and wind turbine type. In other words, damage information is one of the important bases for simulation analysis, but damage images alone are usually insufficient to complete the simulation. Users also need to supplement necessary operating parameters through natural language interaction to improve the accuracy of simulation input.

[0079] Therefore, damage information provides the facts of the on-site damage, constituting the object and background for user-initiated interactions. User natural language input expresses how the user hopes the platform will utilize this damage information—whether it's for explanation, further parameter supplementation, or initiating simulation verification. The equipment identifies these requests through its natural language understanding module and converts them into a standardized format before passing them to the subsequent analysis and simulation modules. Based on this, users submit typhoon disaster assessment requests for wind turbine units via natural language, such as specifying the typhoon path, wind speed, turbine type, and damage data. The equipment parses these natural language inputs, identifying the task requirements and key parameters.

[0080] Leveraging the semantic understanding capabilities of large models, the user interface can perform semantic analysis on the input statements to identify the user's specific needs and transform them into a standardized format that the platform can process for subsequent analysis and simulation.

[0081] For example, let's take a user's input into the device: "What kind of typhoon is Typhoon Mangkhut, and what impact will it have on a certain bay wind farm? To determine whether the blade cracks and tower damage on site are consistent with this typhoon, please help me with a disaster analysis." Figure 4 The detailed process of "extracting disaster analysis simulation parameters from user interaction" is shown below: First, the user interface receives the natural language input and uses the semantic understanding capabilities of the large model to parse the statement. The device recognizes that the input simultaneously includes knowledge-based question-and-answer requirements, impact analysis requirements, and simulation verification requirements. Subsequently, the device extracts explicit parameters from the natural language, including the typhoon object, target wind farm, damage phenomena, and analysis objectives. Since wind load simulations typically require parameters such as wind speed, wind direction, typhoon path, and turbine type, the device further checks whether the current input is missing key simulation parameters. If any are missing, the necessary information is supplemented by prompting the user or calling relevant data interfaces.

[0082] After parameter extraction and completion, the device converts the user's raw natural language into a standardized task format that can be processed internally, including task type, target wind farm, typhoon event, damage location of concern, analysis target, and simulation parameters. Then, the information analysis and task identifier classifies the tasks: those concerning typhoon characteristics and general impacts are identified as inference tasks, where the inference agent generates explanatory answers by combining model knowledge and hybrid retrieval results. Those concerning "determining whether the on-site damage matches typhoon conditions" are identified as tool invocation tasks, using the Model Context Protocol (MCP) to call the wind load simulator and related data interfaces, utilizing typhoon path, wind speed, wind direction, turbine type, and damage information to conduct structural dynamic response simulation. After simulation, the device compares the simulation results with the on-site damage information to determine whether the difference between the on-site damage characteristics and the typhoon condition simulation results meets the preset consistency criteria. This achieves a complete process from natural language input to semantic understanding, parameter extraction, standardized representation, and simulation analysis invocation.

[0083] B. Multimodal data upload and processing.

[0084] Users can also upload other types of data through this module, such as image data and meteorological data. This data may include images of wind turbines, on-site photos, meteorological records, etc., and the equipment can obtain information from multiple data sources.

[0085] All uploaded data is formatted and standardized through the preprocessing module in the user interface. For uploaded image data, the device automatically extracts relevant damage features using image processing technologies such as the YOLO model, and uses these features as input for subsequent analysis. The device automatically identifies and organizes this data, providing the necessary information for subsequent tasks. The device intelligently understands user needs, processes them quickly, and provides feedback, ensuring the smoothness and accuracy of the entire disaster assessment process.

[0086] (3) Information analysis and task identification device.

[0087] The information analysis and task identifier is configured to invoke the inference agent to generate an answer corresponding to the user's instruction when the classification task is an inference task. This module parses the natural language instructions input by the user, identifies the task requirements, and classifies the task according to the requirements. Classification tasks mainly include inference tasks and tool invocation tasks. Inference tasks are transmitted to the inference agent, which generates an initial answer by accelerating the inference framework. Tool invocation tasks use the Model Context Protocol (MCP) to invoke external and internal tools (such as the Zwind simulation model, meteorological data interface, etc.) to complete the simulation task.

[0088] (4) Wind power load simulator.

[0089] The wind power load simulator is configured to invoke an integrated wind power load simulation tool when the task is classified as a tool invocation task. Based on a standardized disaster analysis simulation parameter set, it performs structural dynamic response simulation and control strategy evaluation of offshore wind turbines under typhoon conditions. This module invokes the integrated wind power load simulation tool through a model context protocol to simulate the structural dynamic response of offshore wind turbines under extreme weather conditions such as typhoons, and outputs at least one simulation result, including the maximum equivalent stress at the tower base, tower base bending moment, blade root load, tower top displacement, and damage response index. The device can also perform simulations under multiple combinations of yaw angles and pitch angles, extracting the maximum equivalent stress at the tower base corresponding to each combination. The pitch angle with the lowest maximum equivalent stress at the tower base and meeting the preset stress safety margin requirement is determined as the recommended pitch angle, thus establishing a mapping relationship between yaw angle and recommended pitch angle. When the difference between the simulation results and the field measured data meets the preset consistency criteria, the device obtains the simulation analysis results; when the difference does not meet the preset consistency criteria, the device triggers a closed-loop feedback mechanism, corrects the simulation parameters, and re-performs the simulation.

[0090] Unlike traditional wind turbine simulation tools, such as Figure 5 As shown, the integrated wind power load simulation tool used in this application can connect to multiple tool terminals (such as wind load terminal, wave and current load terminal, soil-structure interaction terminal, control terminal, and structure terminal). It combines wind turbine load simulation, structural response, and control strategy evaluation, filling the gap in existing simulation tools in integrated design optimization and typhoon-resistant control strategies. In particular, it provides an effective solution in hydrodynamic load calculation and load response and structural status evaluation of wind turbine units under typhoon conditions.

[0091] In some embodiments, the control strategy evaluation includes determining a recommended pitch angle strategy under typhoon conditions based on simulation results of different combinations of typhoon wind speed, yaw angle, and pitch angle. Specifically, the device can extract the mapping relationship between the yaw angle and the recommended pitch angle based on the maximum equivalent stress at the tower base under multiple typhoon conditions, and use this mapping relationship as part of the control strategy recommendation for typhoon conditions. In yaw failure scenarios, the mapping relationship can be represented as a nonlinear optimal control trajectory that varies with the yaw angle, used to provide a control strategy reference for typhoon early warning and post-disaster assessment. In some simulation samples, the nonlinear optimal control trajectory may include W-shaped trajectory features, but these W-shaped trajectory features are only used to describe the trend of the recommended pitch angle changing with the yaw angle and do not constitute a limitation on the scope of protection of this application.

[0092] The wind power load simulator can comprehensively evaluate the behavior of wind turbines under extreme weather conditions, and through precise dynamic modeling and adaptive control strategies, it provides a scientific basis for typhoon disaster assessment and repair strategies, ensuring the stability and accuracy of the platform under disaster conditions.

[0093] (5) Result feedback device.

[0094] The results feedback unit is configured to form a closed-loop feedback based on simulation results, and generate a typhoon disaster and disaster resistance assessment report for offshore wind turbines based on at least one of the following: damage identification results, structural dynamic response results, comparison results of on-site measured data, disaster resistance evaluation indicators, and typhoon operating condition control strategy recommendations. This module is used to compare and analyze simulation results with on-site data to verify whether the simulation results meet preset consistency criteria. If the difference between the simulation results and the on-site measured data does not meet the preset consistency criteria, the equipment triggers a feedback mechanism, adjusts the simulation parameters, and re-performs the simulation until the difference between the simulation results and the on-site measured data meets the preset consistency criteria.

[0095] Throughout the process, simulation results are continuously compared and corrected with field feedback data. After each feedback correction, simulation parameters are updated based on field data to improve the consistency between simulation results and field observation data. This closed-loop feedback process enables the use of field observation data for simulation parameter correction and provides data support for typhoon disaster and disaster resilience assessment reports.

[0096] Based on the above description, the device provided in this application is essentially a closed-loop platform that integrates a large language model and a wind power load simulation tool, aiming to solve the technical problems of low evaluation efficiency and poor accuracy caused by the lack of integration, low level of intelligence and insufficient real-time performance in the prior art.

[0097] As an optional implementation, the offshore wind turbine typhoon disaster and disaster resistance assessment equipment provided in this application is packaged as a mobile application, desktop client software or browser program, and supports a distributed deployment architecture for interactive display on smart terminals and model inference and simulation calculation on cloud or edge servers.

[0098] As an optional implementation, the damage detector may include a data collection and annotation module, a dataset partitioning module, a YOLO model training module, and a Model Context Protocol (MCP) invocation module. The data collection and annotation module is configured to collect a large amount of wind turbine image data and form the basis of the training set through annotation. The dataset partitioning module, connected to the data collection and annotation module, is configured to partition the collected damage image data into a training set, validation set, and test set in a 7:2:1 ratio. The YOLO model training module, connected to the dataset partitioning module, is configured to use the YOLO model for object detection, extracting features from images through a convolutional neural network (CNN) to quickly identify and locate target regions in the images. The Model Context Protocol (MCP) invocation module, connected to the YOLO model training module, is configured to invoke the trained YOLO damage detection model via MCP, working collaboratively with other modules within the device.

[0099] As an optional implementation, the data collection and annotation module of the damage detector is further optimized. This module is configured to collect a large amount of wind turbine image data, including different types of damage such as burns, cracks, deformation, and peeling. This image data primarily comes from on-site photography and publicly available datasets. The collected image data is annotated, labeling the damage areas and types within the images to form the basis of the training set. It is precisely this refined data collection and annotation strategy that enables the damage detection module of this system to achieve higher accuracy and robustness.

[0100] As an optional implementation, the user interrogator may further include a natural language understanding module and a multimodal data upload and processing module. The natural language understanding module is configured to parse user-input natural language commands, identifying task requirements and key parameters. The multimodal data upload and processing module is connected to the natural language understanding module and is configured to receive other types of data uploaded by the user, such as image data and meteorological data, and perform formatting and standardization processing through the platform's preprocessing module. For uploaded image data, the device automatically extracts relevant damage features using image processing technologies such as the YOLO model, and uses these features as input for subsequent analysis.

[0101] As an optional implementation, the Natural Language Understanding (NLP) module and the Multimodal Data Upload and Processing (MDP) module of the Interrogator are further optimized. The NLP module is configured to parse user-inputted NLP commands, identifying task requirements and key parameters. Leveraging the semantic understanding capabilities of a large model, this module can perform semantic analysis on the input statements to identify the user's specific needs and convert them into a standardized format that the platform can process for subsequent analysis and simulation. The MDP module receives other types of data uploaded by the user and formats and standardizes them using the platform's preprocessing module. For uploaded image data, the platform automatically extracts relevant damage features using image processing technologies such as the YOLO algorithm and uses these features as input for subsequent analysis. The device can intelligently understand user needs, process them quickly, and provide feedback, ensuring the smoothness and accuracy of the entire disaster assessment process. It is precisely because of the adoption of advanced NLP technology and multimodal data processing strategies that the system's Interrogator has a higher level of intelligence and a better user experience.

[0102] As an optional implementation, the information analysis and task recognizer may further include an inference task processing module, a tool invocation task processing module, and a task classification and scheduling module. The inference task processing module is configured to transmit inference tasks to the inference agent, generating preliminary answers through an accelerated inference framework. The tool invocation task processing module is connected to the inference task processing module and is configured to invoke external and internal tools (such as the Zwind simulation model, meteorological data interfaces, etc.) via the Model Context Protocol (MCP) to complete simulation tasks. The task classification and scheduling module is connected to both the inference task processing module and the tool invocation task processing module. This module is configured to identify task requirements based on user-input natural language commands, classify tasks according to requirements, and schedule the corresponding processing modules.

[0103] As an optional implementation, the inference task processing module and tool invocation task processing module of the information analysis and task recognizer are further optimized. The inference task processing module is configured to transmit inference tasks to the inference agent and generate preliminary answers through an accelerated inference framework. This framework employs parallel computing and optimization algorithms to improve inference speed and accuracy. The tool invocation task processing module invokes external and internal tools (such as the Zwind simulation model, meteorological data interfaces, etc.) through the Model Context Protocol (MCP) to complete simulation tasks. The MCP protocol defines the communication rules and data formats between modules, ensuring the efficiency and stability of tool invocation. The task classification and scheduling module identifies task requirements based on the natural language instructions input by the user, classifies tasks according to requirements, and schedules the corresponding processing modules. In some implementations, the task classification and scheduling module can determine task processing priorities based on task type, data integrity, and user request order, improving the overall efficiency of the system. It is precisely because of the optimized inference task processing and tool invocation strategies that the information analysis and task recognizer of this system has higher processing capacity and flexibility.

[0104] As an optional implementation, the wind power load simulator can further include a self-developed integrated analysis theory and design software module and a data interface module. The self-developed integrated analysis theory and design software module is configured to combine wind turbine load simulation, structural response, and control strategy evaluation, filling the gaps in existing simulation tools regarding integrated design optimization and typhoon-resistant control strategies. This module can comprehensively evaluate the behavior of wind turbines under extreme weather conditions and provide a scientific basis for typhoon disaster assessment and repair strategies through accurate dynamic modeling and adaptive control strategies. The data interface module is connected to the self-developed integrated analysis theory and design software module. The data interface module is configured to receive simulation parameters from the information analysis and task identifier and transmit the simulation results to the result feedback unit.

[0105] As an optional implementation, the self-developed integrated analysis theory and design analysis software module and data interface module of the wind power load simulator are further optimized. The self-developed integrated analysis theory and design analysis software module combines wind turbine load simulation, structural response, and control strategy evaluation, filling the gaps in existing simulation tools regarding integrated design optimization and typhoon-resistant control strategies. This module adopts an integrated aerodynamic-hydraulic-geodynamic-multibody-servo control analysis theory, enabling comprehensive evaluation of wind turbine behavior under extreme weather conditions. Through precise dynamic modeling and adaptive control strategies, it provides a scientific basis for typhoon disaster assessment and repair strategies. The data interface module receives simulation parameters from the information analysis and task identifier and transmits the simulation results to the result feedback unit. This module uses a standardized data interface to ensure the accuracy and efficiency of data transmission. It is precisely because of the adoption of the self-developed integrated analysis theory and standardized data interface that the wind power load simulator of this system has higher evaluation accuracy and compatibility.

[0106] As an optional implementation, the result feedback unit may further include a simulation result comparison and analysis module, a feedback mechanism triggering module, and a closed-loop feedback control module. The simulation result comparison and analysis module is configured to compare and analyze the simulation results with field data to verify the accuracy of the simulation results. The feedback mechanism triggering module is connected to the simulation result comparison and analysis module and is configured to automatically trigger the feedback mechanism and adjust the simulation parameters if an inconsistency is found between the prediction and the field data. The closed-loop feedback control module is connected to the feedback mechanism triggering module and the wind power load simulator and is configured to update the simulation parameters based on the field data to ensure the accuracy and timeliness of the typhoon disaster assessment, thus forming a closed-loop feedback mechanism.

[0107] Based on the above description, compared with the prior art, this application has at least the following advantages: 1. Achieving Integrated Closed-Loop Analysis: This application enables full-process automation from damage detection to disaster assessment. By integrating five core modules (damage detector, user interface, information analysis and task identification, wind load simulator, and result feedback), and closely combining simulation with on-site data feedback, it achieves integrated typhoon disaster assessment. This eliminates the need to switch between multiple tools, reduces manual operation, improves overall work efficiency, and supports full-process automation from damage detection to disaster assessment, solving the problem of low assessment efficiency caused by a lack of integration in existing technologies. Furthermore, by integrating large language models and wind load simulation tools, this application provides an intelligent, integrated, and unified solution for wind turbine disaster assessment, breaking through the fragmented tools and high-barrier operations of traditional technologies. The integrated process improves the efficiency and accuracy of assessment, filling the gap in integrated design optimization and typhoon-resistant control strategies in existing simulation tools. It enables complete automation of the entire process from damage detection to disaster assessment, providing data support for the assessment of the disaster resistance capabilities and control strategy decisions of offshore wind turbines in typhoon disaster environments.

[0108] 2. Improved Accuracy and Real-Time Performance of Disaster Assessment: This application, through a closed-loop feedback mechanism and real-time correction, can dynamically adjust simulation parameters, improving the accuracy of disaster assessment under extreme weather conditions such as typhoons and overcoming the shortcomings of insufficient real-time performance in existing technologies. By comparing with field data, simulation parameter correction can be triggered based on preset consistency criteria, improving the consistency between assessment results and field observation data.

[0109] 3. Reduced operational barriers and improved efficiency: Users can complete disaster assessment tasks for wind turbines simply by receiving and parsing natural language commands. This application automatically parses and executes the tasks, significantly reducing the technical barrier to disaster assessment for wind turbines, improving operational convenience, and solving the problem of low intelligence in existing technologies. Furthermore, this application not only performs real-time corrections based on field data but also provides efficient disaster assessment and prediction through intelligent reasoning and task recognition. Its adaptive feedback mechanism can quickly adapt and adjust the simulation process during extreme weather changes such as typhoons, ensuring real-time and efficient disaster response.

[0110] 4. Improved adaptability and intelligence: By integrating large language models and simulation tools, this application realizes an intelligent and integrated disaster assessment scheme, which not only improves the assessment accuracy, but also enhances the adaptability of the equipment, enabling it to handle various typhoon paths and wind turbine configurations.

[0111] 5. Through multimodal data uploading and processing, it supports information acquisition from multiple data sources, improves the comprehensiveness and accuracy of data utilization, and solves the problem of poor data mobility in existing technologies.

[0112] 6. By using the Model Context Protocol (MCP) to call tools, collaborative work between modules can be achieved, improving the overall efficiency and stability of the equipment and overcoming the shortcomings of existing technologies in terms of collaboration difficulties.

[0113] 7. By comparing and analyzing simulation results and using a feedback mechanism, the assessment results can be corrected based on on-site data feedback, providing a basis for typhoon disaster decision-making, improving the timeliness of typhoon disaster assessment, and solving the problem of delayed feedback in existing technologies.

[0114] 8. By accelerating the inference framework and task classification and scheduling, the intelligence level of the equipment can be improved, enabling efficient disaster assessment and prediction, and overcoming the lack of intelligence capabilities in existing technologies.

[0115] 9. Through dynamic modeling and control strategy evaluation, calculate disaster resistance evaluation indicators such as maximum equivalent stress at the tower base, structural stress margin, recommended pitch angle, and stress reduction rate, to provide data support for typhoon disaster assessment, disaster resistance judgment, and control strategy selection.

[0116] As an scalable implementation, without changing the core principle of this application which focuses on typhoon disaster and disaster resilience assessment, the integrated wind power load simulation tool can also be extended to seismic conditions. For seismic conditions, the device can calculate corresponding structural response indicators based on seismic excitation parameters, unit operating status parameters, and tower base structural response results, and generate control strategy recommendations under seismic conditions. This extended implementation is only used to illustrate that the integrated wind power load simulation tool has multi-extreme condition extension capabilities and does not constitute a limitation on the main process of typhoon disaster and disaster resilience assessment in the claims.

[0117] In one exemplary embodiment, a computer device is provided, which may be a server or a terminal, and its internal structure diagram may be as follows. Figure 6As shown, the computer device includes a processor, memory, input / output (I / O) interfaces, and a communication interface. The processor, memory, and I / O interfaces are connected via a device bus, and the communication interface is also connected to the device bus via the I / O interfaces. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores operating devices, computer programs, and a database. The internal memory provides an environment for the operation of the operating devices and computer programs stored in the non-volatile storage media. The database stores data on typhoon disaster and disaster resilience assessment of offshore wind turbines. The I / O interfaces are used for information exchange between the processor and external devices. The communication interface is used for communication with external terminals via a network connection. When the computer program is executed by the processor, it implements a method for assessing typhoon disaster and disaster resilience of offshore wind turbines.

[0118] Those skilled in the art will understand that Figure 6 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer equipment to which the present application is applied. Specific computer equipment may include, for example, [the following is a list of possible additional structures]. Figure 6 The diagram shows more or fewer components, or combinations of certain components, or different component arrangements.

[0119] In one exemplary embodiment, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the above-described method embodiments.

[0120] In one exemplary embodiment, a computer-readable storage medium is provided storing a computer program that, when executed by a processor, implements the steps in the above-described method embodiments.

[0121] In one exemplary embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above-described method embodiments.

[0122] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data must comply with relevant regulations.

[0123] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments described above. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (RRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM).

[0124] The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.

[0125] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0126] This document uses specific examples 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 methods and core ideas of this application. Furthermore, those skilled in the art will recognize that, based on the ideas of this application, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A method for assessing the typhoon disaster and disaster resistance capabilities of offshore wind turbines, characterized in that, include: Based on multimodal data from offshore wind turbines, damage features are extracted using computer vision and target detection algorithms to generate structured damage information; The structured damage information includes at least one of the following: damaged component identification, damage category, damage location, damage scale, severity level, detection confidence level, data source, acquisition time, and associated image number; Based on user commands, the semantic understanding capabilities of a large language model are combined to perform intent recognition and key parameter extraction, resulting in a classification task and a standardized disaster analysis simulation parameter set. The standardized disaster analysis simulation parameter set includes explicit parameters in the user commands, damage state parameters determined by structured damage information, and environmental condition parameters supplemented through interactive questioning or external data interfaces. User commands are categorized into tasks; when the categorized task involves disaster simulation analysis, a standardized disaster analysis simulation parameter set is generated based on user commands, structured damage information, and environmental condition data. The integrated wind power load simulation tool is then called to perform structural dynamic response simulation under typhoon conditions, and the simulation results are compared with the field measured data. When the difference between the simulation results and the actual measured data does not meet the preset consistency criteria, the simulation parameters are corrected and the simulation is repeated. When the difference between the simulation results and the field measured data meets the preset consistency criteria, the disaster resistance evaluation index is calculated based on the simulation analysis results, and an assessment report on typhoon disaster and disaster resistance of offshore wind turbines is generated. The disaster resistance evaluation index includes at least one of the following: maximum equivalent stress at the tower base, stress safety margin, recommended pitch angle under typhoon conditions, or stress reduction rate relative to the benchmark control strategy. The assessment report includes at least one of the following: damage identification results, structural dynamic response results, comparison results of field measured data, disaster resistance evaluation index, and typhoon condition control strategy recommendations.

2. The method for assessing typhoon disaster and disaster resistance capabilities of offshore wind turbines according to claim 1, characterized in that, Damage features are extracted using multimodal data from offshore wind turbines and computer vision and target detection algorithms. Generate structured damage information, including: The YOLO model is used to locate the target and identify the initial damage category in the multimodal data, thereby obtaining the initial damage area and category. Using a visual analysis model or a large visual model, the severity level of the damage is determined, semantic annotation is performed, and optional treatment suggestions are generated based on the initial damage area and category to obtain the structured damage information; wherein, the treatment suggestions are generated according to the damage category, severity level, and preset operation and maintenance rules.

3. The method for assessing typhoon disaster and disaster resistance capabilities of offshore wind turbines according to claim 1, characterized in that, The process of obtaining the user instruction includes: Obtain the user's voice commands; The voice command is amplified and denoised to obtain the processed voice command. Call the automatic speech recognition service to convert the processed voice commands into text; The text is then cleaned and standardized to produce a standardized text. The standardized text is semantically parsed, intent-classified, and key parameter extracted using a large language model to generate the user instructions.

4. The method for assessing typhoon disaster and disaster resistance capabilities of offshore wind turbines according to claim 1, characterized in that, The process of extracting key parameters based on user commands and leveraging the semantic understanding capabilities of a large language model to obtain a standardized disaster analysis simulation parameter set includes: Based on the semantic understanding capability of the large language model, explicit parameters in the user instructions are extracted, and missing key parameters in the simulation are detected based on structured damage information. The missing key parameters are completed by interactive questioning or calling external data interfaces to obtain the completed parameters. The explicit parameters, the damage state parameters determined by the structured damage information, and the supplementary parameters are standardized and normalized to generate the standardized disaster analysis simulation parameter set.

5. The method for assessing typhoon disaster and disaster resistance capabilities of offshore wind turbines according to claim 1, characterized in that, The integrated wind power load simulation tool uses the integrated coupling analysis theory of aerodynamics-hydraulics-earth motion-multibody-servo control for simulation.

6. The method for assessing typhoon disaster and disaster resistance capabilities of offshore wind turbines according to claim 1, characterized in that, Processing the classification task includes: When the classification task is a reasoning task, the reasoning agent is invoked to generate the answer corresponding to the user instruction; When the classified task is a tool invocation task, the integrated wind power load simulation tool is invoked to simulate and evaluate the structural dynamic response of offshore wind turbines under typhoon conditions based on the standardized disaster analysis simulation parameter set, and to obtain simulation results. By comparing and analyzing the simulation results with the field measurement data, the comparison results are obtained. Determine whether the comparison result exceeds a preset threshold, and obtain the determination result; When the judgment result is yes, the closed-loop feedback mechanism is triggered to correct the simulation parameters and re-perform the simulation. When the judgment result is negative, the simulation analysis result is obtained.

7. The method for assessing typhoon disaster and disaster resistance capabilities of offshore wind turbines according to claim 1, characterized in that, The disaster resistance evaluation indicators include at least one of the following: maximum equivalent stress at the tower base, stress safety margin, damage severity level, recommended pitch angle under typhoon conditions, control strategy recommendations corresponding to the yaw angle, and stress reduction rate relative to the baseline control strategy. The stress safety margin, also known as the structural stress margin, is used to characterize the margin of the offshore wind turbine structure response from the preset allowable stress threshold under typhoon conditions. The stress safety margin is determined based on the difference or ratio between the preset allowable stress threshold and the maximum equivalent stress at the bottom of the tower. The preset allowable stress threshold is determined based on at least one of the following: the yield strength of the tower material, the limit specified in the structural design code, or the preset safety factor. The baseline control strategy includes at least one of the following: fixed pitch angle strategy, shutdown feathering strategy, or preset default control strategy; The stress reduction rate is determined based on the difference between the maximum equivalent stress at the bottom of the tower under the baseline control strategy and the maximum equivalent stress at the bottom of the tower under the recommended control strategy.

8. A device for assessing the typhoon disaster and disaster resistance capabilities of offshore wind turbines, characterized in that, include: Damage detectors are used to extract damage features and generate structured damage information based on multimodal data from offshore wind turbines using computer vision and target detection algorithms. The structured damage information includes at least one of the following: damaged component identification, damage category, damage location, damage scale, severity level, detection confidence level, data source, acquisition time, and associated image number; The user interactor is used to perform intent recognition and key parameter extraction based on user commands and the semantic understanding capabilities of a large language model, to obtain a classification task and a standardized disaster analysis simulation parameter set. The standardized disaster analysis simulation parameter set includes explicit parameters in user commands, damage state parameters determined by structured damage information, and environmental condition parameters supplemented by interactive questioning or external data interfaces. Information analysis and task identifier, used to classify user commands into tasks; The wind load simulator is used to generate a standardized disaster analysis simulation parameter set based on user instructions, structured damage information and environmental condition data when the classification task involves disaster simulation analysis, and then call the integrated wind load simulation tool to perform the simulation analysis. The result feedback device is used to form a closed-loop feedback based on the simulation analysis results, and to trigger simulation parameter correction and re-simulation when the difference between the simulation results and the field measured data does not meet the preset consistency criteria. The result feedback device is also used to generate an assessment report on typhoon disaster and disaster resistance capabilities of offshore wind turbines based on at least one of the answers corresponding to the user instructions, simulation analysis results, disaster resistance evaluation indicators, and typhoon operating condition control strategy suggestions. The damage detector, the user interactor, the information analysis and task identifier, the wind power load simulator, and the result feedback device are all integrated and deployed on the same platform, and run in series with the intelligent agent workflow through a unified software architecture.

9. A computer device, comprising: A memory, a processor, and a computer program stored in the memory and capable of running on the processor, characterized in that the processor executes the computer program to implement the method for assessing the typhoon disaster and disaster resistance capabilities of offshore wind turbines as described in any one of claims 1-7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the computer program implements the method for assessing the typhoon disaster and disaster resistance capabilities of offshore wind turbines as described in any one of claims 1-7.

Images