Offshore wind turbine generator set one-machine one-file checking method based on deep learning

Abstract

The invention belongs to the technical field of offshore wind power file checking, and particularly provides a one-machine one-file checking method for an offshore wind power unit based on deep learning, and the method comprises the following steps: 1, data collection and preprocessing; 2, constructing a deep neural network model, and checking the archive data preprocessed in the step 1; and step 3, generating a checking report. According to the method, the checking work of the fan file can be automatically completed, and the problems of tedious manual checking and low efficiency in the prior art are solved.

Description

Technical Field

[0001] The present invention belongs to the technical field of offshore wind power file verification, and specifically relates to a one-machine-one-file verification method for offshore wind turbines based on deep learning. Background Art

[0002] With the rapid development of the wind power industry, especially in offshore wind power, the operation and maintenance of wind turbines is becoming increasingly complex. Within a wind farm, each wind turbine requires a detailed file (known as a "one-turbine-one-file" system). These files cover all aspects of the turbine's lifecycle, including installation, operation, maintenance, overhaul, and fault records. While this "one-turbine-one-file" management model can improve wind farm management accuracy, the existing data review and comparison process still relies on manual labor, which is labor-intensive and prone to errors.

[0003] Currently, wind farm "one-machine-one-file" data typically exists in various formats, including spreadsheets, PDF files, and video surveillance. Manual verification of this data is not only time-consuming but also prone to omissions and errors, impacting wind farm operational efficiency and timely turbine maintenance. Therefore, an automated, intelligent verification method is urgently needed to improve the accuracy and efficiency of data review. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to provide a one-machine-one-file calibration method for offshore wind turbines based on deep learning, which automatically completes the calibration of wind turbine files and solves the problems of tedious and inefficient manual calibration in the existing technology.

[0005] To solve the above technical problems, the technical solution adopted by the present invention is: a method for verifying the performance of an offshore wind turbine generator system based on deep learning, comprising the following steps: Step 1: Data collection and preprocessing; Step 2: Build a deep neural network model to verify the archival data preprocessed in step 1; Step 3: Generate a verification report.

[0006] In a preferred solution, in step 1, data collection is to obtain archival data of the wind turbine from a wind farm database or a file storage system.

[0007] In a preferred solution, the archival data of the wind turbine includes structured data, unstructured data and video monitoring data, the structured data includes data in Excel and CSV formats; the unstructured data includes PDF files and pictures.

[0008] In a preferred solution, in step 1, data preprocessing includes removing irrelevant data, fixing format errors, and data standardization.

[0009] In a preferred solution, in step 2, the verification of the archival data includes structured data verification, unstructured data verification and video surveillance data verification.

[0010] In a preferred embodiment, the operation method of structured data verification is: constructing a deep neural network model, analyzing tabular data, and automatically checking the accuracy and consistency of the data; the deep neural network model adopts a multi-layer feedforward neural network structure, including an input layer, several hidden layers and an output layer, the input layer receives the standardized tabular data features, the hidden layer uses the ReLU activation function to enhance the nonlinear modeling capability, and the output layer uses the softmax or sigmoid function to classify or score the verification results.

[0011] In a preferred embodiment, the unstructured data verification operation method is to use image recognition technology and natural language processing methods to extract key information from the document and compare it with the standard list to identify potential inconsistencies or errors.

[0012] In a preferred solution, the method for verifying unstructured data includes the following steps: First, optical character recognition technology is used to convert PDF files and image content into editable text. Then, the named entity recognition method in natural language processing is used to extract the key fields of time, location, equipment number, and event type. Next, the extracted fields are matched and semantically compared with the wind farm standard file template to determine whether there are missing items, logical inconsistencies, or data anomalies. Finally, the comparison results are fed back to the verification module for the generation of the verification report in step three.

[0013] In a preferred solution, the operation method for verifying the video surveillance data is: automatically analyzing the video surveillance data of the wind turbine generator set, and the automatic analysis includes the following steps: first, performing frame extraction and image preprocessing on the video data; then, using a deep convolutional neural network to detect and classify targets in key frames; then, using a timing analysis model to identify the operating status of the equipment; finally, comparing the identified operating status with the corresponding event records in the archive to verify whether there are any unregistered events or recording errors.

[0014] In a preferred embodiment, in step three, the verification report includes a text description, a data comparison chart, and a visual image of the analysis results.

[0015] The present invention provides a deep learning-based one-machine-one-gear calibration method for offshore wind turbines, which has the following beneficial effects: 1. Automation and intelligence: Through deep learning technology, the system can automatically perform data verification, reduce manual participation, and avoid omissions or errors that may occur during manual review.

[0016] 2. High efficiency: The traditional manual verification process takes a lot of time. The present invention can complete the verification of a large amount of archival data in a relatively short time, greatly improving work efficiency.

[0017] 3. Accuracy: Utilizing deep neural networks and natural language processing technology, it can accurately identify potential issues in archives, avoiding possible omissions during manual review.

[0018] 4. Strong adaptability: Whether it is structured data or unstructured data (such as PDF, pictures, videos), the method of the present invention can effectively process them and adapt to the verification needs of various data formats.

[0019] 5. Scalability: This technical solution can be flexibly integrated into the existing wind farm operation and maintenance management platform, and has good scalability and can be applied to more wind farms and equipment management systems. DETAILED DESCRIPTION

[0020] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0021] It should be noted that the terms "first," "second," and the like in the description and claims of the present invention are used to distinguish similar objects and are not necessarily used to describe a specific order or precedence. It should be understood that the terms used in this manner are interchangeable where appropriate so that the embodiments of the present invention described herein can be implemented in an order other than those described herein. In addition, the terms "including" and "having," as well as any variations thereof, are intended to cover non-exclusive inclusions. For example, a process, method, system, product, or apparatus comprising a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units that are not explicitly listed or that are inherent to these processes, methods, products, or apparatus.

[0022] Example 1: A method for verifying one offshore wind turbine at one level based on deep learning, comprising the following steps: Step 1: Data collection and preprocessing.

[0023] Obtain wind turbine archival data from the wind farm database or file storage system. This data includes but is not limited to structured data (such as Excel spreadsheets and CSV files) and unstructured data (such as PDF documents, scanned documents, video surveillance, etc.).

[0024] That is, the wind turbine's archival data includes structured data, unstructured data, and video monitoring data. The structured data includes data in Excel and CSV formats; the unstructured data includes PDF files and pictures.

[0025] Data preprocessing includes removing irrelevant data, fixing format errors, and data standardization to ensure the quality of input data.

[0026] The data standardization includes: unifying the time format (for example, unifying it to the "YYYY-MM-DD" format), unifying the numerical units (for example, unifying the temperature to ℃ and the wind speed to m / s), normalizing the field names (for example, unifying fields such as "maintenance records" and "maintenance status" to "maintenance records"), processing missing values ​​(for example, using interpolation, mean filling, or deep model prediction filling), and encoding text fields (for example, using one-hot encoding or word vector embedding) to ensure the consistency and processability of the data in subsequent analysis.

[0027] Step 2: Build a deep neural network model to verify the archival data preprocessed in step 1, including structured data verification, unstructured data verification, and video surveillance data verification.

[0028] The operation method of the structured data verification is: building a deep neural network (DNN) model, analyzing tabular data, such as maintenance records, repair cycles, equipment replacement history, etc., and automatically checking the accuracy and consistency of the data. For example, automatically matching data items through algorithms to check whether there are missing items, erroneous items, etc.

[0029] The deep neural network model adopts a multi-layer feedforward neural network structure, including an input layer, several hidden layers and an output layer. The input layer receives the standardized table data features, the hidden layer uses the ReLU activation function to enhance the nonlinear modeling capability, and the output layer uses the softmax or sigmoid function to classify or score the verification results. The model can be trained based on historical verification samples. During the training process, cross entropy is used as the loss function, and the Adam optimizer is used to improve the convergence speed. Through this model, the system can identify problems such as missing items, abnormal field values, and logical contradictions between fields in the table, and realize automatic verification of structured data.

[0030] The unstructured data verification method involves extracting key information from unstructured data such as PDF files and images using image recognition technology and natural language processing methods for proofreading. For wind turbine maintenance records and inspection reports, key information in the text is automatically identified and compared against standard checklists to identify potential inconsistencies or errors.

[0031] The specific steps include: first, using OCR (optical character recognition) technology to convert PDF files and image content into editable text; then, using the named entity recognition (NER) method in natural language processing (NLP) to extract key fields such as time, location, equipment number, event type, etc.; then, the extracted fields are matched and semantically compared with the wind farm standard file template to determine whether there are missing items, logical inconsistencies or data anomalies; finally, the comparison results are fed back to the verification module and used for report generation.

[0032] The operation method of the video monitoring data verification is: automatically analyzing the video monitoring data of the wind turbine generator set, identifying the operating status, equipment status and other information of the unit through computer vision technology, and matching and verifying it with archival data.

[0033] Automatic analysis includes the following steps: first, frame extraction and image preprocessing (such as noise reduction and enhancement) of the video data; then, a deep convolutional neural network (such as YOLO or ResNet) is used to detect and classify targets in key frames (such as wind turbine blades, hubs, towers, etc.); then, a time series analysis model (such as LSTM or 3D-CNN) is used to identify the equipment operating status (such as start, stop, fault shutdown, etc.); finally, the identified operating status is compared with the corresponding event records in the archive to check whether there are any unregistered events or recording errors.

[0034] Step 3: Generate a verification report.

[0035] During the verification process, the system automatically generates a verification report based on the analysis results. This report lists potential issues, errors, or omissions found, and provides correction suggestions. The report includes not only a text description but also data comparison charts or visualizations of the analysis results, allowing technicians to quickly identify issues.

[0036] The system can also provide real-time feedback based on user needs to help managers track the progress of problem resolution.

[0037] The automatic verification method of the present invention can be seamlessly connected to the existing wind farm operation and maintenance management system, and the verification results can be fed back to the system through an API interface or a data synchronization mechanism for automatic updating and maintenance.

[0038] Users can set the verification time period (such as monthly, quarterly, etc.) as needed. The system will automatically verify the files according to the set frequency, avoiding human intervention and improving work efficiency.

[0039] This paper proposes an automated verification method that combines deep neural networks (DNN) and natural language processing (NLP) technologies, addressing the high cost and low efficiency of existing manual verification. It can process a variety of data formats (such as structured data, PDFs, images, and videos), automatically extracting and verifying unstructured data through computer vision and NLP. It automatically generates verification reports and provides real-time feedback to managers, helping them identify and correct data issues and improving management efficiency. Regular automated verification is supported, eliminating the tedious process of manual verification, and verification cycles can be flexibly set according to user needs.

[0040] Example 2: According to the operating method in Example 1, 20 wind turbines in offshore wind farm A were selected as the research objects. Structured (CSV), unstructured (PDF and images), and video monitoring data for nearly one year were extracted from its archive system, and automated verification was performed using the verification system proposed in the present invention.

[0041] Structured data is collated using a deep neural network (DNN) model. Unstructured data uses OCR and natural language processing to extract key information before comparison. Video data uses YOLOv5 and LSTM models to identify device operating status and compare it with recorded data. Table 1 shows the accuracy of automatic collation for various data types. This data demonstrates that the proposed collation method can replace manual collation.

[0042]

[0043] It will be easily understood by those skilled in the art that the above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A deep learning-based one-machine-one-gear calibration method for offshore wind turbines, characterized in that: The following steps are involved: Step 1: Data collection and preprocessing; Step 2: Build a deep neural network model to verify the archival data preprocessed in step 1; Step 3: Generate a verification report.

2. A method for verifying one offshore wind turbine generator system based on deep learning according to claim 1, characterized in that: In the step 1, data collection is to obtain the wind turbine archive data from the wind farm database or file storage system.

3. The method for verifying one offshore wind turbine generator system based on deep learning according to claim 2, characterized in that: The wind turbine archive data includes structured data, unstructured data and video monitoring data. The structured data includes data in Excel and CSV formats; the unstructured data includes PDF files and pictures.

4. The method for verifying one offshore wind turbine generator system based on deep learning according to claim 1, characterized in that: In the step 1, data preprocessing includes removing irrelevant data, fixing format errors, and data standardization.

5. The method for intelligent verification of two tickets of offshore wind turbines based on deep learning according to claim 1 is characterized in that: In the step 2, the verification of the archive data includes structured data verification, unstructured data verification and video surveillance data verification.

6. The method for intelligent verification of two tickets of offshore wind turbines based on deep learning according to claim 5 is characterized in that: The operation method of the structured data verification is as follows: constructing a deep neural network model, analyzing the tabular data, and automatically checking the accuracy and consistency of the data; the deep neural network model adopts a multi-layer feedforward neural network structure, including an input layer, several hidden layers and an output layer, the input layer receives the standardized tabular data features, the hidden layer uses the ReLU activation function to enhance the nonlinear modeling capability, and the output layer uses the softmax or sigmoid function to classify or score the verification results.

7. The method for intelligent verification of two tickets of offshore wind turbines based on deep learning according to claim 5 is characterized in that: The operation method of the unstructured data verification is: using image recognition technology and natural language processing methods to extract key information from documents, and compare it with the standard list to find potential inconsistencies or errors.

8. The method for intelligent verification of two tickets of offshore wind turbines based on deep learning according to claim 7 is characterized in that: The operation method of the unstructured data verification The following steps are involved: First, optical character recognition technology is used to convert PDF files and image content into editable text. Then, the named entity recognition method in natural language processing is used to extract the key fields of time, location, equipment number, and event type. Next, the extracted fields are matched and semantically compared with the wind farm standard file template to determine whether there are missing items, logical inconsistencies, or data anomalies. Finally, the comparison results are fed back to the verification module for the generation of the verification report in step three.

9. The method for intelligent verification of two tickets of offshore wind turbines based on deep learning according to claim 5 is characterized in that: The operation method of the video surveillance data verification is: automatically analyzing the video surveillance data of the wind turbine generator set, and the automatic analysis includes the following steps: first, frame extraction and image preprocessing of the video data; then, using a deep convolutional neural network to detect and classify targets in key frames; then, using a time series analysis model to identify the equipment operating status; finally, comparing the identified operating status with the corresponding event records in the archive to verify whether there are any unregistered events or recording errors.

10. The method for intelligent verification of two tickets of offshore wind turbines based on deep learning according to claim 1, characterized in that: In step three, the verification report includes a text description, a data comparison chart, and a visual image of the analysis results.

Images