A method for collecting and generating a report of factory witnessing data of a power transformation device

Abstract

The application discloses a kind of substation equipment factory witnessed data acquisition and report generation method, it is related to electric power equipment detection technical field.The method is through industrial-grade intelligent hardware platform and special software system cooperation, realize equipment identity multimodal accurate identification, acceptance data structured acquisition, multimedia evidence security solidification and standardization report one-key generation.Its core steps include: industrial-grade hardware adaptation and initialization, multimodal equipment identity matching, visual acceptance guidance and data acquisition, space-time watermark multimedia evidence generation, acceptance data encryption processing, standardization report automatic assembly.The application solves the problem of non-standard data collection, unreliable evidence chain and low report generation efficiency in traditional substation equipment factory witnessed, realizes paperless, standardization and traceability of acceptance work, greatly improves the efficiency of acceptance and data credibility, and is suitable for various substation equipment factory acceptance scenarios.

Description

Technical Field

[0001] This invention belongs to the field of power equipment testing technology, specifically, it relates to a method for collecting and generating factory witness data for substation equipment. Background Technology

[0002] Factory acceptance testing of power equipment is a crucial step in ensuring the safe and stable operation of the power system and must be strictly carried out in accordance with the relevant acceptance standards of the State Grid Corporation of China. Traditional factory acceptance witnessing work relies on paper-and-pen records, independent camera photography, and subsequent computer data entry and report compilation, which has many drawbacks: First, acceptance data recording is often non-standardized, prone to omissions and errors, and manual entry is inefficient; second, evidence such as on-site photos lacks effective anti-counterfeiting mechanisms, key information such as time and location can be tampered with, and the credibility of the evidence chain is insufficient; third, the preparation of acceptance reports relies on manual data integration and formatting, taking several hours or even longer, and the format is difficult to standardize to meet State Grid standards; fourth, unstructured acceptance detail documents are difficult to access quickly, requiring acceptance personnel to repeatedly consult paper or electronic documents, affecting acceptance efficiency.

[0003] While existing technologies have made some attempts at digitalization, they mostly use consumer-grade hardware, which cannot adapt to complex industrial environments; equipment identification methods are limited and matching accuracy is low; and there is a lack of an integrated data collection and report generation process, failing to fundamentally solve the pain points of the traditional model. Therefore, there is an urgent need for a method for collecting and generating data for the factory witnessing of power equipment that balances environmental adaptability, data reliability, and process efficiency.

[0004] In view of this, the present invention is proposed. Summary of the Invention

[0005] To solve the above-mentioned technical problems, the basic concept of the technical solution adopted by the present invention is as follows:

[0006] A method for collecting and generating factory witness data for power equipment includes the following steps:

[0007] Step S1: Industrial-grade hardware platform adaptation and initialization: Build an industrial-grade hardware platform containing a high-performance camera, a high-precision GPS module, a 4G communication module and a long-lasting battery, complete hardware anti-electromagnetic interference and wide-temperature adaptation debugging, and pre-install a dedicated software system and digital acceptance template library.

[0008] Step S2: Multimodal device identification and template matching: Obtain device identification information through three methods: QR code or barcode scanning, OCR nameplate recognition, or manual selection from a structured list, and call the corresponding acceptance template based on the preset mapping relationship;

[0009] Step S3: Visualized Acceptance Guidance and Data Collection: The acceptance clauses are transformed into a visual UI interface to guide the acceptance personnel to select "qualified", "unqualified" or "not involved" results, and simultaneously collect multimedia data such as equipment appearance and key parameters. The multimedia data includes photos and videos.

[0010] Step S4: Spatiotemporal watermark multimedia evidence generation: Collect GPS positioning data and UTC time data from the hardware platform, and automatically embed an immutable watermark at the moment of multimedia data acquisition;

[0011] Step S5: Acceptance data encryption processing: The acceptance result data and watermarked multimedia evidence are encrypted and stored locally, supporting offline operation. After the network is restored, they are synchronized to the backend management system through an HTTPS / VPN secure channel.

[0012] Step S6: Automatic generation of standardized reports: Based on the template engine developed by Apache POI, the system automatically aggregates acceptance data, multimedia evidence, acceptance details and overall conclusions, and generates a daily acceptance report in Word or PDF format that conforms to the State Grid's standards within 30 seconds.

[0013] As a preferred embodiment of the present invention, the adaptation and debugging of the industrial-grade hardware platform in step S1 includes: stability testing in a wide temperature range of -20℃ to 60℃, under strong electromagnetic interference and humid environments.

[0014] In a preferred embodiment of the present invention, the priority of multimodal device identification in step S2 is as follows: QR code / barcode scanning is given priority, OCR nameplate recognition is used as an auxiliary method, and manual selection from a structured list is used as a fallback solution to achieve comprehensive device identification.

[0015] In a preferred embodiment of the present invention, the watermark in step S4 includes the project name, equipment ID, acceptor, time, and geographical coordinates.

[0016] In a preferred embodiment of the present invention, when the standardized report is automatically generated in step S6, the number of qualified and unqualified items is automatically counted, and the text, pictures and signature information are formatted according to a preset format to ensure that the report is 100% compliant with the latest standard template format of the State Grid Corporation of China.

[0017] As a preferred embodiment of the present invention, it also includes a data traceability step: establishing an association index between acceptance data and multimedia evidence, supporting the query of corresponding acceptance records and original evidence by keywords such as project name, equipment ID, and acceptance time.

[0018] As a preferred embodiment of the present invention, the visual acceptance guidance in step S3 also includes a pop-up window for the interpretation of acceptance terms. When the acceptance personnel have questions about the terms, they can view detailed explanations to ensure the accuracy of the acceptance judgment.

[0019] In a preferred embodiment of the present invention, the local encrypted storage in step S5 adopts the AES-256 encryption algorithm, and a data integrity verification mechanism is adopted during the data synchronization process to avoid data loss or tampering.

[0020] In a preferred embodiment of the present invention, the OCR nameplate recognition employs a deep learning-based text detection and recognition model, wherein text detection uses the YOLOv5 algorithm, text recognition uses the CRNN model, and the recognition probability is calculated using the Softmax function, as shown in the formula: in, Indicates in Predicting character categories from input images probability; It is a category The score K is the total number of character categories. When the recognition confidence is higher than the threshold... If the condition is met, output the recognition result; otherwise, trigger manual selection from the structured list.

[0021] In a preferred embodiment of the present invention, the spatiotemporal watermark embedding employs a discrete wavelet transform domain watermarking algorithm. The specific steps include: performing a two-level DWT decomposition on the acquired image to obtain low-frequency sub-band coefficients, and embedding the watermark information into the low-frequency sub-band. The embedding formula is as follows: in, It is the coefficient after embedding the watermark. It is an embedded intensity factor with a value range of 0.01-0.05 to ensure that the watermark is invisible and robust; then, the image is reconstructed by inverse DWT to generate multimedia evidence with spatiotemporal watermark.

[0022] Compared with the prior art, the present invention has the following advantages:

[0023] This invention ensures accurate template matching through multimodal device identification, avoids omissions in acceptance items through visual guidance, and guarantees the immutability of multimedia evidence through spatiotemporal watermarking technology, thereby improving the quality of acceptance data from the source. Based on an industrial-grade hardware platform and after rigorous environmental adaptability testing, it can operate stably in complex industrial scenarios such as strong electromagnetic interference, high and low temperatures, and humidity, meeting the on-site usage requirements of power equipment factory acceptance.

[0024] The specific embodiments of the present invention will now be described in further detail with reference to the accompanying drawings. Attached Figure Description

[0025] In the attached diagram:

[0026] Figure 1 This is a flowchart of a method for collecting and generating factory witness data for power equipment. Detailed Implementation

[0027] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments will be clearly and completely described below with reference to the accompanying drawings. The following embodiments are used to illustrate the present invention.

[0028] The present invention provides a method for collecting and generating factory witness data for power equipment, comprising the following steps:

[0029] Step S1: Industrial-grade hardware platform adaptation and initialization

[0030] An industrial-grade hardware platform was built, incorporating a high-performance camera, a high-precision GPS module, a 4G communication module, and a long-lasting battery. Hardware electromagnetic interference protection and wide-temperature adaptability testing were completed, and a dedicated software system and digital acceptance template library were pre-installed to operate under conditions ranging from -20℃ to 60℃.

[0031] The construction process of the digital acceptance template library is as follows: Natural Language Processing (NLP) is used to parse 28 volumes of the General Management Regulations for Substation Acceptance of Power Grid Companies. A BERT-based semantic understanding model is used to extract structured fields such as acceptance objects, acceptance clauses, qualification standards, and related evidence types to form a template database containing more than 3,000 standard clauses that can be directly called by software.

[0032] Step S2: Multimodal device identification and template matching

[0033] Device identification information is obtained through three methods, and the corresponding acceptance template is called based on the preset mapping relationship: QR code / barcode scanning: the high-performance camera of the hardware platform is called first, and open source libraries such as ZBar / Zxing are used for fast decoding; OCR nameplate recognition: as an auxiliary method, it is enabled when there is no code or the code is damaged; manual selection of structured list: as a backup solution, a drop-down menu with three levels of linkage, namely substation, bay, and equipment type, is provided for users to select.

[0034] The priority for multimodal device identification is as follows: QR code or barcode scanning is given priority, OCR nameplate recognition is used as an auxiliary method, and manual selection from a structured list is used as a fallback solution to achieve comprehensive device identification.

[0035] As a key improvement of this invention, the OCR nameplate recognition adopts a deep learning-based integrated model for text detection and recognition:

[0036] Text detection: Using the YOLOv5 algorithm, the bounding box prediction formula is as follows:

[0037]

[0038]

[0039]

[0040]

[0041] in,( , , ( ) represents the center coordinates and width / height of the prediction box; , , () is the network output; , ) represents the grid coordinates; , () represents the prior frame size; For the Sigmoid function;

[0042] Text recognition: A CRNN (Convolutional Recurrent Neural Network) model is used. It first extracts image feature sequences using a CNN (such as VGG16), then feeds them into a Bi-LSTM (Bidirectional Long Short-Term Memory) network for sequence context modeling, and finally trains using a CTC (Connectivity-Temporal Classification) loss function to recognize text sequences of variable length. The recognition probability is calculated using the Softmax function, with the following formula:

[0043]

[0044] in, Indicates the features of the input image Predict character categories The probability of; K is the score of category i; K is the total number of character categories; in practice, a confidence threshold is set. =0.95. When the model's overall sequence confidence for the identified device model "S11-M-800 / 10" is higher than 0.95, the identification result is output; otherwise, the system automatically triggers and jumps to the structured list manual selection interface to prevent incorrect matching.

[0045] Step S3: Visualized Acceptance Guidance and Data Collection

[0046] Abstract acceptance criteria are transformed into an intuitive, visual UI interface. Each criterion is presented in card format, guiding acceptance personnel to select "qualified," "unqualified," or "not applicable" results by clicking on them. Simultaneously, multimedia data such as equipment appearance and key parameter meter readings are collected from cameras.

[0047] Furthermore, the visual acceptance guidance also includes a pop-up window explaining the acceptance clauses. When acceptance personnel have questions about a clause (such as "circuit breaker mechanical characteristic test"), they can click the "question mark" icon next to it to trigger a pop-up window to view detailed explanations, standard operation example diagrams, and even short videos reviewed by experts, ensuring the accuracy and consistency of acceptance judgments.

[0048] Step S4: Spatiotemporal watermark multimedia evidence generation involves collecting GPS positioning data (latitude and longitude coordinates) from the hardware platform and UTC time data synchronized from the network. An immutable digital watermark is automatically embedded at the moment the multimedia data (photo / video frames) is captured.

[0049] The spatiotemporal watermark embedding employs a Discrete Wavelet Transform (DWT) domain watermarking algorithm, and the specific steps include the following:

[0050] S1: Preprocessing: Encode and spread the watermark information W (text containing project name, equipment ID, acceptor, UTC time and geographical coordinates) to generate a binary watermark sequence.

[0051] S2:DWT decomposition: decomposition of the acquired raw image A two-stage discrete wavelet transform is performed to decompose the low-frequency subband (LL), horizontal high-frequency subband (LH), vertical high-frequency subband (HL), and diagonal high-frequency subband (HH). The coefficients of the low-frequency subband with the strongest embedding robustness are selected. .

[0052] S3: Watermark Embedding: The watermark sequence W is embedded into the low-frequency subband using additive rules. The embedding formula is:

[0053] in, It is the coefficient after embedding the watermark. It is an embedded strength factor with a value range of 0.01-0.05 to ensure that the watermark is invisible and robust; then, the image is reconstructed by inverse DWT to generate multimedia evidence with spatiotemporal watermark. It can achieve the best robustness while ensuring that the watermark is invisible (PSNR > 40dB), and can resist attacks such as JPEG compression, noise interference and slight geometric shearing.

[0054] S4: Reconstruction: Use modified coefficients Perform inverse discrete wavelet transform (IDWT) to reconstruct and generate the final multimedia evidence image with spatiotemporal watermark. .

[0055] Step S5: Acceptance data encryption processing. All data, including structured acceptance results and watermarked multimedia evidence, are encrypted and stored locally using the AES-256 encryption algorithm. The encryption mode uses GCM mode, which provides not only confidentiality but also data integrity authentication. The entire process supports offline operation to ensure normal operation in remote areas or factory areas with poor signal. After network restoration, the encrypted data packets are synchronized to the backend management system via HTTPS protocol based on TLS 1.3 or an IPSec VPN secure channel. During synchronization, the SHA-256 hash algorithm is used for data integrity verification, generating a data digest, which is uploaded along with the data packets. Receipt is only confirmed after the backend verification passes, effectively preventing data loss or tampering during transmission.

[0056] Step S6: Standardized report is automatically generated. Based on an intelligent template engine developed using Apache POI, it automatically aggregates all acceptance data, multimedia evidence, electronic signatures, and overall conclusions. The report generation engine has a built-in rule-based automatic statistics module that can count the number of qualified and unqualified items in real time and calculate the pass rate. Through precise API control of Apache POI, text, images, and signature information are automatically formatted according to the preset format of the latest State Grid standard template, ensuring that the generated Word or PDF format acceptance daily report is 100% compliant with the specifications.

[0057] It also includes data traceability steps: In the backend management system, an index linking acceptance data and multimedia evidence is established. Users can perform combined searches using multiple keywords such as project name, equipment ID, and acceptance time to quickly locate and retrieve the corresponding complete acceptance records and all original evidence, forming a complete and credible chain of evidence.

[0058] Example 1

[0059] This embodiment applies to the factory acceptance testing of a 110kV transformer, and uses the method described in this invention for data acquisition and report generation. The specific steps are as follows:

[0060] Industrial-grade hardware platform adaptation and initialization: A brand-name industrial-grade tablet PC was selected, integrating a high-precision GPS module, a 13-megapixel camera, and a dedicated QR code scanning module. Stability testing was completed under strong electromagnetic interference conditions to ensure a battery life of up to 10 hours. A dedicated software system was pre-installed, and 28 State Grid acceptance detail templates were imported into the structured parsing library. Hardware and software integration and debugging were completed, and the device was ready to use upon startup.

[0061] Multimodal equipment identification and template matching: When the acceptance personnel scan the QR code on the transformer nameplate, the system matches the "110kV transformer factory acceptance template" in milliseconds and automatically loads the corresponding acceptance clauses, totaling 86 items.

[0062] Visualized acceptance guidance and data collection: The software interface displays the acceptance clauses one by one, such as "winding DC resistance measurement" and "insulating oil dielectric loss test". The acceptance personnel check "qualified" according to the on-site test results and take 12 photos of the transformer appearance, bushing installation and other parts. The system automatically optimizes the image clarity.

[0063] Spatiotemporal watermark multimedia evidence generation: When taking a photo, the system automatically embeds watermark information (Project Name: XX110kV Transformer Factory Acceptance; Equipment ID: BYS-2025-001; Acceptance Person: [Name]; Time: 2025-10-15 14:32:18; Latitude and Longitude: 125.36°E, 46.58°N), and the watermark cannot be tampered with.

[0064] Acceptance data encryption: All acceptance results and photos are stored locally in an encrypted manner. After acceptance is completed (network restored), they are synchronized to the State Grid Xing'an Power Supply Company's back-end management system through a VPN secure channel.

[0065] Standardized report automatically generated: When the acceptance personnel click "Generate Report", the system generates a daily acceptance report in PDF format within 25 seconds, automatically counting 86 qualified items and 0 unqualified items, and attaching 12 watermarked photos and acceptance details. The format fully complies with the latest State Grid standards.

[0066] Data traceability and management: The back-end management system records complete data for this acceptance, and supports quick querying of acceptance reports, original photos and acceptance process records by device ID "BYS-2025-001".

[0067] The entire acceptance process, from data collection to report generation, took only 8 minutes, saving 2.5 man-hours compared to the traditional method. The report format was standardized, and the evidence chain was complete and traceable.

[0068] Example 2

[0069] This embodiment is applied to the factory acceptance scenario of 35kV switchgear. Due to the damage to the equipment nameplate, it is impossible to scan the code. Therefore, a combination of OCR recognition and manual selection is used. The specific steps are as follows:

[0070] Industrial-grade hardware platform adaptation and initialization: Using the same industrial-grade hardware platform as in Example 1, after initialization, the digital acceptance template library is loaded.

[0071] Multimodal equipment identification and template matching: After the acceptance personnel failed to scan the QR code on the switchgear nameplate, they activated the OCR recognition function, took a photo of the damaged nameplate, and the system identified the equipment model "KYN28-12". The system matched the "35kV switchgear factory acceptance template" through keywords. After verification, the acceptance personnel confirmed that the template was correctly selected.

[0072] Visualized acceptance guidance and data collection: Follow the interface guidance to complete the selection of 68 acceptance clauses, and take 8 photos of the internal wiring of the switch cabinet and the status of the circuit breaker. Two of the photos automatically enabled HDR imaging due to low light to ensure image clarity.

[0073] Spatiotemporal watermark multimedia evidence generation: All 8 generated photos are automatically embedded with complete watermark information, with an indoor positioning accuracy of 18 meters, which meets the acceptance requirements.

[0074] Acceptance data encryption: All data collection is completed while the network is offline, and the data is automatically synchronized to the backend system after the network is restored, ensuring that the data transmission is undamaged and unaltered.

[0075] Standardized report automatically generated: A daily acceptance report in Word format is generated within 30 seconds, including two non-conformities ("minor damage to cabinet door sealing strip" and "blurred nameplate markings") and corresponding watermarked photos, with automatic annotation of rectification suggestions for non-conformities.

[0076] Data traceability and management: The back-end system records information on non-conformities, facilitating subsequent tracking and rectification. Acceptance data can be quickly retrieved using keywords such as project time and equipment model.

[0077] The entire acceptance process took 10 minutes, saving approximately 300 yuan in labor costs compared to the traditional method. The acceptance report was formatted correctly, and the evidence for any non-compliance items was clear and verifiable.

Claims

1. A method for collecting and generating reports on factory-issued witness data for power equipment, characterized in that, Includes the following steps: Step S1: Industrial-grade hardware platform adaptation and initialization: Build an industrial-grade hardware platform containing a high-performance camera, a high-precision GPS module, a 4G communication module and a long-lasting battery, complete hardware anti-electromagnetic interference and wide-temperature adaptation debugging, and pre-install a dedicated software system and digital acceptance template library. Step S2: Multimodal device identification and template matching: Obtain device identification information through three methods: QR code or barcode scanning, OCR nameplate recognition, or manual selection from a structured list, and call the corresponding acceptance template based on the preset mapping relationship; Step S3: Visualized Acceptance Guidance and Data Collection: The acceptance clauses are transformed into a visual UI interface to guide the acceptance personnel to select the results and simultaneously collect multimedia data such as equipment appearance and key parameters. The multimedia data includes photos and videos. Step S4: Spatiotemporal watermark multimedia evidence generation: Collect GPS positioning data and UTC time data from the hardware platform, and automatically embed an immutable watermark at the moment of multimedia data acquisition; Step S5: Acceptance data encryption processing: The acceptance result data and watermarked multimedia evidence are encrypted and stored locally, supporting offline operation. After the network is restored, they are synchronized to the backend management system through an HTTPS / VPN secure channel. Step S6: Automatic generation of standardized reports: Based on the template engine developed by Apache POI, the system automatically aggregates acceptance data, multimedia evidence, acceptance details and overall conclusions, and generates a daily acceptance report in Word or PDF format that conforms to the State Grid's standards within 30 seconds.

2. The method for collecting and generating reports on factory-issued witness data for power equipment according to claim 1, characterized in that, The adaptation and debugging of the industrial-grade hardware platform in step S1 includes: stability testing in a wide temperature range of -20℃ to 60℃, under strong electromagnetic interference and humid environments.

3. The method for collecting and generating reports on factory-issued witness data for power equipment according to claim 1, characterized in that, In step S2, the priority of multimodal device identification is as follows: QR code / barcode scanning is given priority, OCR nameplate recognition is used as an auxiliary method, and manual selection from a structured list is used as a fallback solution to achieve comprehensive device identification.

4. The method for collecting and generating reports on factory-issued witness data for power equipment according to claim 1, characterized in that, The watermark mentioned in step S4 includes the project name, equipment ID, acceptor, time, and geographical coordinates.

5. The method for collecting and generating reports on factory-issued witness data for power equipment according to claim 1, characterized in that, When the standardized report is automatically generated in step S6, the number of qualified and unqualified items is automatically counted, and the text, images and signature information are formatted according to the preset format to ensure that the report is 100% compliant with the latest standard template format of the State Grid Corporation of China.

6. The method for collecting and generating reports on factory-issued witness data for power equipment according to claim 1, characterized in that, It also includes data traceability steps: establishing an index linking acceptance data and multimedia evidence, supporting the query of corresponding acceptance records and original evidence by keywords such as project name, equipment ID, and acceptance time.

7. The method for collecting and generating reports on factory-issued witness data for power equipment according to claim 1, characterized in that, Step S3, the visual acceptance guidance, also includes a pop-up window explaining the acceptance terms. When the acceptance personnel have questions about the terms, they can view detailed explanations to ensure the accuracy of the acceptance judgment.

8. The method for collecting and generating reports on factory-issued witness data for power equipment according to claim 1, characterized in that, In step S5, the local encrypted storage uses the AES-256 encryption algorithm, and a data integrity verification mechanism is used during data synchronization to prevent data loss or tampering.

9. The method for collecting and generating reports on factory-issued witness data for power equipment according to claim 3, characterized in that, The OCR nameplate recognition employs a deep learning-based text detection and recognition model. Text detection uses the YOLOv5 algorithm, and text recognition uses a CRNN model. The recognition probability is calculated using the Softmax function, with the formula as follows: in, Indicates in Predicting character categories from input images probability; It is a category The score K is the total number of character categories. When the recognition confidence is higher than the threshold... If the condition is met, output the recognition result; otherwise, trigger manual selection from the structured list.

10. The method for collecting and generating reports on factory-issued witness data for power equipment according to claim 4, characterized in that, The spatiotemporal watermark embedding employs a discrete wavelet transform domain watermarking algorithm. The specific steps include: performing a two-level DWT decomposition on the acquired image to obtain low-frequency sub-band coefficients, and embedding the watermark information into the low-frequency sub-band. The embedding formula is as follows: in, It is the coefficient after embedding the watermark. It is an embedded intensity factor with a value range of 0.01-0.05 to ensure that the watermark is invisible and robust; then, the image is reconstructed by inverse DWT to generate multimedia evidence with spatiotemporal watermark.

Images