Field operation data analysis method and system

Abstract

The invention provides a field operation data analysis method and system, and belongs to the technical field of information, and the method comprises the steps: inputting the power distribution equipment type, equipment installation position and planned operation time basic information of an operation task into a unified to-be-examined operation database according to the to-be-examined operation work order information of the power distribution field operation; collecting an examination work video image, and analyzing the operation standardability of an operator; evaluating the work efficiency and quality, and generating rectification suggestions; and comparing historical data to identify anomalies, and optimizing an operation process to realize effective integration and analysis of power distribution field operation data.

Description

Technical Field

[0001] The present invention relates to the field of information technology, and in particular to a method and system for analyzing field operation data. Background Art

[0002] During the on-site review of power distribution operations, reviewers are faced with a massive amount of operation data and information, which are usually scattered in different systems and documents, lacking effective integration and analysis methods. Due to the complexity and uncertainty of the on-site operation environment of power distribution, reviewers need to dynamically adjust and optimize the operation plan and safety measures according to the actual situation of the distribution site. This requires reviewers to quickly and comprehensively grasp various data and information related to the operation, and to be able to promptly discover problems and risks, which requires video image recording to analyze the data. However, the current situation of data dispersion and lack of standardized operation methods for analysis operators makes it difficult for reviewers to carry out their work efficiently and safely, which not only affects the quality and efficiency of the review work, but also increases the safety hazards of on-site distribution operations. How to achieve effective integration and analysis of operation data and improve the work efficiency and quality of reviewers has become a technical problem that needs to be solved urgently. Summary of the invention

[0003] In order to solve the above technical problems, the present invention provides a field operation data analysis method and system.

[0004] The present application embodiment discloses a method for analyzing field operation data, which mainly includes:

[0005] According to the work order information of the power distribution field operation to be reviewed, the basic information of the power distribution equipment type, equipment installation location, and planned operation time of the operation task is entered into the unified database of operations to be reviewed;

[0006] Through the electronic tags or QR codes of the power distribution equipment, the model specifications, rated parameters, operating status, and maintenance record equipment information of the equipment are collected to improve the equipment information in the database of the operation to be reviewed;

[0007] According to the improved database of pending review operations, the identity information and qualifications of the operators involved in the pending review operations are obtained and associated with the corresponding operation tasks;

[0008] Go to the corresponding work site according to the corresponding task, and collect and review the temperature, humidity, and noise environmental conditions of the work node environment in real time according to different work task types and work environment requirements. If the environmental indicators exceed the safety threshold of the work node task type, an early warning will be issued to the reviewer. If the work node environmental indicators meet the safety threshold, the review work process nodes will be collected in the form of video and image data;

[0009] Analyze the collected video and image data of the review operation process nodes to determine whether the operators are operating in accordance with the operation node operating procedures, including whether they are using qualified operation tools and equipment and whether they comply with the operation safety protection requirements, and record the judgment results in the operation process node records;

[0010] Based on the recorded operation process nodes, the completion efficiency of the operation tasks is compared and analyzed according to the planned completion time and the actual completion time of the operation tasks. The operation quality is scored in combination with the operation process node records to form an operation result evaluation. For the operation tasks that have not passed the review, the reasons for their failure are analyzed, and rectification suggestions are generated based on the equipment failure history and equipment environmental conditions;

[0011] Based on the evaluation of the operation results, the data of this review operation will be compared and analyzed with the historical data of the same type of operation to identify abnormal fluctuations in operation efficiency and quality. If an abnormality is found, the person responsible for equipment management that caused the abnormality will be traced, and the operation plan will be adjusted, the operation process nodes will be optimized, and the data analysis of the on-site review operation nodes will be completed.

[0012] The present application also discloses a field operation data analysis system, including:

[0013] Information entry module, used to enter information on power distribution equipment type, installation location and planned operation time;

[0014] Equipment information collection module, used to collect equipment model, rated parameters, operating status and maintenance records;

[0015] The personnel information association module is used to obtain and associate the identity information and qualifications of the operating personnel with the operating tasks;

[0016] Environmental monitoring module, used to collect and monitor the environmental conditions of the operation nodes in real time to ensure safe operation;

[0017] The operation process analysis module is used to analyze the collected video images and determine whether the operation is compliant;

[0018] The work result evaluation module is used to compare and analyze work efficiency and quality, generate rectification suggestions and optimize.

[0019] The technical solution provided by the embodiment of the present invention may have the following beneficial effects:

[0020] The present invention discloses a method for analyzing on-site operation data. The method establishes a database of operations to be reviewed by collecting operation work order information and equipment data; uses the Internet of Things technology to monitor the operation environment in real time to ensure safety; collects and reviews operation video images, and analyzes the standardization of operation by operators; evaluates operation efficiency and quality, and generates rectification suggestions; compares historical data to identify anomalies and optimizes the operation process. The present invention realizes intelligent management of the entire process of power distribution operations, and improves the accuracy and efficiency of operation review. Through a data-driven approach, the present invention can promptly discover problems in operations and make improvements, effectively improve the safety and quality of power distribution operations, and provide guarantees for the stable operation of the power system. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 The figure is a flow chart of the field operation data analysis method of the present invention. DETAILED DESCRIPTION

[0022] The technical solution of the present invention will be clearly and completely described below in conjunction with the embodiments. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0023] like Figure 1 The on-site operation data analysis method of this embodiment may specifically include:

[0024] Step S101, based on the work order information of the power distribution field operation to be reviewed, the basic information of the power distribution equipment type, equipment installation location, and planned operation time of the operation task is entered into a unified database of operations to be reviewed.

[0025] Acquire the image information of the work order, extract text data from the image information, and identify key information including the type of power distribution equipment, equipment installation location and planned operation time; send a similarity calculation instruction to the preset equipment type vocabulary for the power distribution equipment type, and the preset equipment type vocabulary calculates the similarity value using the edit distance algorithm based on the similarity calculation instruction; receive a coordinate conversion instruction with equipment installation location information, and the coordinate conversion instruction is generated based on the equipment installation location information; determine whether the key information complies with the specified field mapping rules, and if so, write the key information into the database of jobs to be reviewed.

[0026] Specifically, the work order of the electronic document or paper scan is scanned by optical character recognition technology, the text information on the work order is extracted, and the key fields such as the type of power distribution equipment, the equipment installation location, and the planned operation time are identified by named entity recognition technology. The identified power distribution equipment type is matched with the preset equipment type vocabulary, and the similarity is calculated using the edit distance algorithm. The equipment type with the highest similarity is selected as the final result. At the same time, the equipment installation location information is processed using the geographic information system tool and converted into standardized longitude and latitude coordinates. For the extracted planned operation time, regular expressions are used to format it and uniformly converted into the standard time format of "yyyy-mm-ddhh:mm:ss". If the time information is incomplete, the missing time information is inferred based on the work calendar and standard operation time. According to the pre-established data model, the normalized information such as the type of power distribution equipment, the installation location coordinates, and the planned operation time are written into the database of pending operations according to the specified field mapping rules, and a unique identifier is generated for each record using the UUID algorithm. The entered data is checked for integrity and consistency to verify whether the equipment type exists in the preset word library, whether the installation location coordinates are within the valid range, and whether the operation time conforms to the specified format and logic, ensuring that the data quality meets the requirements of the database of pending operations. When processing the work order information, the electronic document or paper scan is first processed using optical character recognition technology to extract text information. For example, for a scanned PDF work order, the Tesseract OCR engine is used for text recognition, and the recognition accuracy can reach more than 95%. Subsequently, a named entity recognition model based on conditional random fields is used to identify key information from the extracted text. The model has been trained on a large number of distribution work orders, and the recall rate F1 score can reach 0.92. For the identified distribution equipment type, such as "10kV transformer", the edit distance algorithm is used to match it with the standard terms in the preset word library. If the edit distance is less than 2, the match is considered successful. Equipment installation location information, such as "Dianli Road Substation in Chengbei District", is converted into latitude and longitude coordinates (A, B) by calling the Amap API. The scheduled operation time is formatted using the regular expression "(\d{4})-(\d{2})-(\d{2})(\d{2}):(\d{2}):(\d{2})". If the time information is incomplete, such as only the date "2024-08-15", it is automatically completed to "2024-08-1509:00:00" according to the work calendar. The processed information is written to the MongoDB database. The data model contains fields such as device type, installation location (latitude and longitude), scheduled operation time, and work order ID, which is generated using the UUIDv4 algorithm.Finally, the entered data is verified, such as checking whether the equipment type is in the preset vocabulary, whether the coordinates are within the valid range, which includes the longitude range of A1°E to A2°E and the latitude range of A3°N to A4°N), and whether the operation time is within the next 30 days.

[0027] Step S102, collect the model specifications, rated parameters, operating status, maintenance record equipment information of the distribution operation equipment through the electronic tag or QR code of the distribution operation equipment, and improve the equipment information of the operation database to be reviewed.

[0028] Use an RFID reader to scan the electronic tag of the power distribution equipment, or use an optical scanner to read the QR code of the equipment to obtain the unique identifier and basic information of the equipment. According to the unique identifier and basic information of the equipment, obtain a processable data format. Query the equipment information that matches the unique identifier of the equipment. The equipment information includes model specifications, rated parameters, historical operating status and maintenance records. Compare the processable data format with the historical records in the MySQL relational database. If the similarity is lower than the preset threshold, it is judged as an inconsistent item. Associate the equipment information with the corresponding records in the database of pending operations, update the model specifications, rated parameters, operating status and maintenance records in the database of pending operations. If it is judged to be a new device or the equipment information does not exist in the database, create a new equipment record in the database of pending operations.

[0029] Specifically, an RFID reader is used to scan the electronic tag of the distribution operation equipment, or an optical scanner is used to read the QR code to obtain the unique identifier and basic information of the equipment stored in the tag, and the obtained information is converted into a processable data format through Base64 decoding or JSON parsing. According to the decoded unique identifier of the equipment, the matching detailed equipment information, including model specifications, rated parameters, historical operating status and maintenance records, is queried from the MySQL relational database. The information obtained from the electronic tag or QR code is compared with the historical records in the database through the Levenshtein distance algorithm, and the similarity threshold is set to 0.9 to judge the consistency and validity of the information. If the similarity is lower than the threshold, it is marked as an inconsistent item. The verified equipment information is associated with the corresponding records in the database of the operation to be reviewed, and the model specifications, rated parameters, operating status and maintenance records of the equipment are updated or supplemented. For the marked inconsistent items, they are recorded in the log file, and the database is updated according to the preset rules. If it is judged to be a new device or equipment information that does not exist in the database, a new equipment record is automatically created. Finally, the information update timestamp is recorded in the database of the operation to be reviewed to complete the supplement and improvement of the equipment information. In the process of collecting information about distribution equipment, first use the ZebraFX9600 fixed RFID reader to scan the electronic tag, or use the ZebraDS9908 barcode scanner to read the QR code to obtain the unique identifier and basic information of the equipment. For example, the RFID tag may contain

[0030] The JSON data in the format of "{"id":"T001","type":"transformer","voltage":"10kV"}" is parsed by the JSON.parse() function. For the QR code, the Base64-encoded data "VDAwMSx0cmFuc2Zvcm1 lciwxMGtW" may be stored, which is decoded by the atob() function. The decoded data is used as a keyword to execute the SQL query "SELECT*FROM equipment WHERE id='T001'" in the MySQL database to obtain the complete equipment information record. Subsequently, the Levenshtein distance algorithm is used to compare the newly acquired information with the database records, such as the equipment type "transformer" and the "Transformer" stored in the database. The calculated distance is 1 and the similarity is 0.9, which is higher than the preset threshold of 0.85, and it is determined to be consistent. For inconsistent items, such as the voltage value "10kV" and the "11kV" in the database, they are recorded in the log.txt file. According to the preset rules, if the inconsistent items account for less than 20% of the total number of fields, the database is automatically updated; otherwise, it is marked for manual review. For new equipment, if the transformer information with ID "T002" is obtained and does not exist in the database, execute the SQL statement "INSERT INTO equipment(id,type,voltage)VALUES('T002','transformer','10kV')" to create a new record. Finally, use the SQL update statement "UPDATE work_orders SETequipment_info='Updated JSON string',update_time=CURRENT_TIMESTAMP WHEREorder_id='WO001'" to associate the new equipment information with the work order to be reviewed and record the update timestamp.

[0031] Step S103, obtaining the identity information and qualifications of the operators who participated in the current operation to be reviewed according to the improved database of the operation to be reviewed, and associating them with the corresponding operation tasks.

[0032] Obtain a list of personnel identity identifiers related to the job in the job database, and the list of personnel identity identifiers is extracted from the job database to be reviewed; based on the list of personnel identity identifiers, use a data integration tool to connect the job database to be reviewed and the human resources database, and query the identity information and qualification information of the operators from the human resources database; perform text analysis on the historical job records of the operators, extract keywords and construct professional ability vectors, and calculate the professional ability matching score of each person through cosine similarity; calculate the number of completions and success rates of similar job types by the operators in the past six months, and obtain the historical job experience evaluation results; combine the professional ability matching score and the historical job experience evaluation results to generate a comprehensive score, write the comprehensive score back to the job database to be reviewed, and establish the association between personnel information and specific job tasks.

[0033] Specifically, the task information of the job to be reviewed is extracted from the improved job database, and the list of personnel identifiers related to the job is obtained through the SQL query statement "SELECT personnel_id FROM work_orders WHERE order_id = 'WO001'". The data integration tool Talend Open Studio is used to connect the job database and the human resources database. According to the obtained personnel identifier, the detailed identity information and qualification information of each operator are queried, including name, work number, professional category, qualification certificate number and validity period. The TF-IDF algorithm is used to perform text analysis on the historical work records of the operators, extract keywords and construct professional ability vectors, and calculate the degree of matching with the current job task through cosine similarity to obtain the professional ability matching score of each person. The statistical method is used to calculate the number of completions and success rates of similar job types of operators in the past 6 months, and to quantitatively evaluate the historical work experience. The professional ability matching score and the historical work experience evaluation results are combined to generate a comprehensive score, and the SQL update statement "UPDATE work_orders SET

[0034] Personnel_info=JSON_OBJECT('id',personnel_id,'score',comprehens ive_score)WHERE order_id='WO001'",write the analysis results back to the database of pending operations,and establish the association between personnel information and specific operation tasks.In the process of processing personnel information of power distribution operations,firstly,use SQL query "SELECTpersonnel_id FROM work_orders WHERE order_id='WO001'"to extract the list of relevant personnel IDs of pending operation WO001 from the operation database,assuming that three operators with IDs "P001","P002" and "P003" are obtained.Then,use Talend Open Studio data integration tool to connect the operation database and human resources database through JDBC connector,and execute cross-database query "SELECT

[0035] name,employee_number,speciality,certificate_number,expiry_date FROMhr_database.employee_info WHERE id IN('P001','P002','P003')", to obtain detailed information of these three personnel. Then, the TF-IDF algorithm is used to process the historical work records of each personnel. For example, the record of P001 "high-voltage line maintenance, transformer installation, switchboard maintenance" is segmented and weighted to obtain the keyword vector [0.5, 0.3, 0.2]. This vector is calculated with the vector [0.1, 0.8, 0.1] of the current job task description "10kV transformer replacement" for cosine similarity, and the matching degree is 0.78. At the same time, the number of similar operations completed by P001 in the past 6 months is 15 times, with a success rate of 95%. Taking into account the professional ability matching and historical experience, the weighted average method is adopted, with weights of 0.6 and 0.4 respectively, and the comprehensive score of P001 is calculated to be 0.852. Finally, through the SQL update statement "UPDATE work_orders SETpersonnel_info=JSON_OBJECT('id','P001','score',0.852)WHERE order_id='WO001'"Write the result back to the job database to complete the association between personnel information and job tasks. Repeat the above process for P002 and P003 to achieve automatic processing and association of all job personnel information.

[0036] Step S104, go to the corresponding work site according to the corresponding task, and collect and review the temperature, humidity, and noise environmental conditions of the work node environment in real time according to different work task types and work environment requirements; if the environmental indicators exceed the safety threshold of the work node task type, an early warning prompt is issued to the reviewer; if the work node environmental indicators meet the safety threshold, the review work process nodes will be collected in the form of video and image data.

[0037] According to the type of job task, the environmental safety threshold configuration is obtained from the preset MySQL database. The environmental safety threshold configuration includes the upper and lower limits of the three indicators of temperature, humidity, and noise. Sensor nodes are arranged at the job site using temperature and humidity sensors and noise sensors. The sensor nodes collect environmental data once every preset time. The environmental data is analyzed in real time to determine whether the environmental data exceeds the environmental safety threshold. If the environmental data meets the environmental safety threshold for a preset time, the video and image data of the job process node are collected and reviewed. The environmental data, video and image data are associated with the job task ID, and the associated data obtained is encapsulated.

[0038] Specifically, according to the type of work task and the requirements of the work environment, the corresponding environmental safety thresholds are retrieved from the preset MySQL database, including the upper and lower limits of the three indicators of temperature, humidity, and noise, and the threshold configuration of the specific task type is obtained through SQL query statements. Using DHT22 temperature and humidity sensors and BK2155 noise sensors, a sensor node is arranged every 1 meter at the work site, data is collected every 5 seconds, and the encrypted data is transmitted to the central processing unit through the ZigBee wireless communication module. The collected environmental data is analyzed in real time using if-then rules to determine whether it exceeds the preset safety threshold. If it exceeds the threshold, the early warning mechanism is triggered, and the early warning information is pushed to the mobile terminal of the reviewer, while the change trend of the environmental data is continuously monitored. When the environmental indicators meet the safety threshold for 10 minutes, the on-site 4K high-definition camera is started, the minimum resolution is set to 3840x2160, and the frame rate is not less than 30fps. The video and image data of the review work process nodes are collected, and the collected data is transmitted to the cloud storage platform through the 5G network. The real-time environmental data, video and image information are associated with the task ID, encapsulated in JSON format, and stored in the MongoDB database to form a complete review operation record. In the distribution operation environment monitoring, the safety thresholds of specific tasks are first obtained through the MySQL database query statement "SELECT temperature_min, temperature_max, humidity_min, humidity_max, noise_max FROM safety_thresholdsWHERE task_type = 'high-voltage line maintenance'", such as the temperature range of -10℃ to 35℃, the humidity range of 20% to 80%, and the noise upper limit of 85dB. Subsequently, a DHT22 temperature and humidity sensor and a BK2155 noise sensor are arranged every 1 meter in the 10m x 10m operation area, with a total of 100 sensor nodes. Each node collects data every 5 seconds and transmits it to the central processing unit through the ZigBee module with AES-128 bit encryption. The central processing unit uses if-then rules for real-time data analysis, such as "IF temperature<-10ORtemperature>35ORhumidity<20ORhumidity>80ORnoise>85THENtrigger_alert()". If an early warning is triggered, a push message containing the specific exceeded value is immediately sent to the reviewer's mobile terminal. During the continuous monitoring of environmental data, when all indicators remain within the safe range for 10 consecutive minutes, 4 Sony FDR-AX434K cameras are automatically started to shoot the work site from different angles. The video is set to 3840x2160 resolution and 30fps frame rate.The collected video and image data are uploaded to Alibaba Cloud Object Storage Service via 5G network at a rate of 200Mbps. Finally, the environment data, video URL and image URL are associated with the job task ID, such as "HV-REPAIR-20240810-001", encapsulated into JSON format {"task_id":"HV-REPAIR-20240810-001","env_data":[...],"video_urls":[...],"image_urls":[...]}", and stored in the "audit_records" collection of the MongoDB database, completing the automated collection and storage process of the audit job records.

[0039] When collecting video and image data in the review operation process nodes, the motion trajectory of the node operators and equipment is detected and tracked in real time. By analyzing the continuity and integrity of the motion trajectory, it is determined whether there is target loss or trajectory interruption due to occlusion or motion blur during the video and image data collection process, and the image data reliability assessment results are given.

[0040] The target detection algorithm is used to process the video frame sequence to obtain the target bounding box coordinates of the operator and equipment; the detection results are smoothed according to the target bounding box coordinates to obtain the preliminary motion trajectory of the target; the Euclidean distance between adjacent frames in the target bounding box coordinate sequence is calculated, and if the Euclidean distance is greater than the preset dynamic threshold, it is determined to be a trajectory interruption; the occlusions in the video frames are identified, and the overlapping area ratio between the target bounding box and the occlusion is calculated. If the overlapping area ratio exceeds the preset threshold, it is determined to be an occlusion state; the optical flow method is used to calculate the pixel motion speed in the video frame sequence. If the pixel motion speed exceeds the preset threshold, it is determined to be motion blur; based on the determination results of trajectory interruption, occlusion state and motion blur, the weighted average method is used to calculate the data reliability score.

[0041] Specifically, the YOLOv5 target detection algorithm is used to locate the operators and equipment in the video frame in real time, obtain the target bounding box coordinates, and use the Kalman filter to smooth the detection results. The optimal filter parameters are obtained through historical trajectory data training to obtain the preliminary motion trajectory of the target. According to the target bounding box coordinate sequence, the Euclidean distance of the target position between adjacent frames is calculated, and the dynamic threshold is set to 1.5 times the diagonal length of the target bounding box. If the distance is greater than the threshold, it is determined to be a trajectory interruption, and the start time and number of frames of the interruption are recorded. The MaskR-CNN instance segmentation algorithm is used to identify objects that may cause occlusion in the image, and the overlapping area ratio of the target bounding box and the occlusion is calculated. If it exceeds 50%, it is marked as an occlusion state, and the occlusion duration and occlusion area ratio are recorded. The optical flow method is used to calculate the pixel motion speed in the image sequence. When the speed exceeds 20 pixels per frame, it is determined to be motion blur. The number and duration of blur frames are counted, and the trajectory interruption, occlusion and blur are comprehensively considered. The weighted average method is used to calculate the data reliability score, with weights of 0.4, 0.3 and 0.3 respectively. The reliability assessment results are associated with the original video and image data in JSON format and stored in the review operation record collection of the MongoDB database. In the on-site monitoring of power distribution operations, the YOLOv5 target detection algorithm is first used to process the video stream with a resolution of 1920x1080 and 30fps, and the bounding box coordinates of the operator and equipment are detected in each frame. For example, the bounding box of operator A is detected as [x=500, y=300, w=100, h=200]. Subsequently, a Kalman filter is applied, where the process noise covariance Q=0.01 and the measurement noise covariance R=0.1 are obtained by analyzing 1000 frames of historical data. After filtering, a smooth trajectory coordinate sequence is obtained. The Euclidean distance of the target position of adjacent frames is calculated, such as the distance from frame1e to frame2 is 15 pixels. The dynamic threshold is set to √(w^2+h^2)1.5, that is, 212 pixels. If the distance exceeds the threshold, the record is interrupted, such as "interruption starts: frame 125, lasts: 10 frames". Next, the Mask R-CNN algorithm is used to identify potential occluders. For example, if an occluder with an area of ​​5000 square pixels is detected, the intersection of the target and the occluder is calculated. If it exceeds 50% of the target area (in this case, 1000050% = 5000 square pixels), it is marked as occlusion and the "occlusion start time is frame 200, duration is 5 seconds, and occlusion ratio is 60%" is recorded. Then, the FarnebFck optical flow algorithm is applied to calculate the pixel movement speed. When the speed exceeds 20 pixels / frame, it is judged as blur and recorded.

[0042] Blur starts: frame 300, lasts: 8 frames". Finally, the reliability score is calculated according to the formula Score = 0.4*(1-interruption frames / total frames) + 0.3*(1-occlusion time / total time) + 0.3*(1-blurred frames / total frames). For example, score = 0.4*(1-10 / 1000) + 0.3*(1-5 / 60) + 0.3*(1-8 / 1000) = 0.9687. The results are stored in the "audit_records" collection of MongoDB in the format of {"video_id":"V001","rel iabi l ity_score":0.925,"interrupt ions":[...],"occlus ions":[...],"blurs":[...]} to complete the reliability assessment.

[0043] Step S105, analyze the collected video and image data of the review operation process node to determine whether the operator operates in accordance with the operation node operating procedures, including whether qualified work tools and equipment are used and whether the work safety protection requirements are complied with, and record the judgment result in the operation process node record.

[0044] The video frames are segmented to obtain target recognition results including operators, tools and equipment, and the working environment. According to the target recognition results, the position, shape and color feature information of each target are extracted. The feature information is compared with the standard operations stored in the database to determine whether the operation behavior meets the specifications, including extracting the model and specification information of the operating tools and equipment in the video frame. According to the model and specification information, it is matched with the list of qualified tools required by the task to determine whether the tools used meet the requirements, including recording the tool use angle and applied force data. The video frames are analyzed by the OpenPose posture estimation algorithm to obtain the body movement and position data of the operators. The video frames are processed by the deep learning model to identify the wearing of safety protection equipment. According to the safety protection requirements of the operation, it is evaluated whether the operators strictly abide by the safety regulations. The video frame sequence is analyzed in time series to obtain the segmentation results of the operation process. For each time series node, a JSON format record containing the operation behavior, tool use and safety protection status is generated. The JSON format record is compared with the standard operation process to calculate the operation compliance score, tool use correctness and safety protection compliance rate, and the calculation results are recorded in the operation process node record.

[0045] Specifically, MaskR-CNN is used to segment the video frames, identify the operators, tools and working environment, extract the position, shape and color features of each target, and compare the recognition results with the standard operating procedures stored in the MySQL relational database. The database is updated by regularly synchronizing the operating procedures. The YOLOv5 target detection algorithm is used to identify the model and specifications of the operating tools and tools, and the detection results are matched with the list of qualified tools required by the operating task to determine whether the tools used meet the requirements and record the use of the tools, including the angle and force of the tools. The OpenPose posture estimation algorithm is used to analyze the body movements and positions of the operators, and the deep learning model is combined to identify the wearing of safety protection equipment. The safety protection requirements of the operation are compared to evaluate whether the operators strictly abide by the safety regulations, and weight scores are assigned to each key protective equipment such as helmets and safety belts. The operation process is analyzed and segmented according to the long short-term memory network (LSTM), and the operation behavior, tool use and safety protection of each node are compared with the standard operation process to generate node records in JSON format, including operation compliance scores, tool use correctness and safety protection compliance rates. When there is uncertainty in the recognition results, an ensemble learning method is used to integrate the outputs of multiple models to improve the robustness of the system. During the review of distribution operations, the MaskR-CNN model is first used to perform image segmentation on video frames with a resolution of 1920x1080 to identify workers, tools and equipment, and the working environment. For example, the worker is detected at coordinates (500,300) and is wearing blue overalls; the insulating pliers are located at (800,400) and are red. These features are compared with the standard operating procedures stored in the MySQL database, and the database is updated with the latest operating procedures through the ETL tool every week. Next, the YOLOv5 algorithm is applied to identify tools. For example, the insulating pliers model DL-500 are detected with a confidence level of 0.95. The algorithm also analyzes the angle at which the tool is used, which is 45° from the horizontal plane.

[0046] And estimated force, which is determined by the degree of deformation of the tool, about 200N. Then use the OpenPose algorithm to analyze the posture of the operator, such as detecting the positions of the hands (510,320) and (490,320), to determine whether the operating posture is standard. At the same time, the deep learning model identifies safety equipment, gives helmets, safety belts, and insulating gloves weights of 40%, 30%, and 30% respectively, and calculates the protection score. Finally, the LSTM network performs time series analysis on the 30fps video stream, with each 10 seconds as a node, and generates JSON format records:

[0047] {"node_id":1,"timestamp":"2024-08-1014:30:00","compl iance_score":0.85,"tool_usage_correctness":0.92,"safety_equipment_score":0.95}. When the confidence of a certain recognition is lower than 0.6, ensemble learning is started to combine the output results of ResNet50 and Mobi leNetV2 to improve recognition accuracy.

[0048] Step S106: Based on the recorded work flow nodes, the completion efficiency of the work tasks is compared and analyzed according to the planned completion time and the actual completion time of the work tasks. Combined with the work flow node records, the work quality is scored to form a work result evaluation. For work tasks that have not passed the review, the reasons for their failure are analyzed, and rectification suggestions are generated based on the equipment failure history and equipment environmental conditions.

[0049] Obtain the operation process node records from the database. The operation process node records contain the planned completion time and actual completion time of each node. Calculate the time difference percentage based on the planned completion time and the actual completion time, and use the ARIMA model to compare and analyze the completion efficiency of the operation task to obtain the efficiency score. If the efficiency score is lower than the preset threshold, the efficiency of the operation task is judged to be unqualified. The weighted average method is used to calculate the operation quality score for the operation process node records, the correctness of tool use, and the safety protection compliance rate. If the operation quality score is lower than the preset threshold, the quality of the operation task is judged to be unqualified. The random forest algorithm is used to analyze and determine the failure causes of the failed operation tasks. The random forest algorithm associates the operation process node records with the equipment failure history data to identify potential equipment problems and environmental factors. According to the potential equipment problems and environmental factors, the failure causes are classified into three categories: operation errors, equipment failures, and environmental impacts. Extract rectification experience from historical successful cases and generate targeted rectification suggestions based on the failure cause categories.

[0050] Specifically, the operation process node records are extracted from the database, the planned completion time and actual completion time of each node are obtained, the time difference percentage is calculated, and the task complexity coefficient is considered. The ARIMA model is used to compare and analyze the completion efficiency of the operation tasks to obtain the efficiency score. According to the preset scoring criteria, combined with the operation standardization in the operation process node records, its weight is 40%, the correctness of tool use, its weight is 30%, and the safety protection compliance rate, its weight is 30%, the weight is 30%, the weight is 40%, and the weight is 50%. The weighted average method is used to calculate the operation quality score and form an operation result evaluation report. For the operation tasks that failed the review, the random forest algorithm is used to analyze the failure reasons, the operation process node records are associated with the equipment failure history data, potential equipment problems and environmental factors are identified, and the failure reasons are classified into three categories: operation errors, equipment failures, and environmental impacts. Based on the association rule mining algorithm Apriori, the rectification experience is extracted from the historical successful cases, combined with the specific failure cause category of the current operation task, targeted rectification suggestions are generated, and the suggestions are stored in the knowledge base. The knowledge base content is updated weekly using the incremental update mechanism. The D3.js visualization library is used to transform the rectification suggestions into interactive flowcharts, highlighting the key rectification steps, which are easy for operators to understand and execute. In the process of power distribution operation evaluation, the operation process node records are first extracted from the MySQL database. For example, the planned completion time of node A is 2024-08-10 14:00, the actual completion time is 2024-08-10 14:30, and the calculation time difference percentage is 25%. Considering the task complexity coefficient of 1.2, the ARIMA (1,1,1) model is used to analyze the completion efficiency, and the efficiency score is 85 points. Subsequently, according to the preset scoring criteria, the operation quality score is calculated as 95 points (weight 40%) for operation standardization, 88 points (weight 30%) for tool use correctness, and 92 points (weight 30%) for safety protection compliance rate. The weighted average is 92.1 points for the total operation quality score. For tasks that failed the review, such as a transformer maintenance job, the random forest algorithm is used with a tree depth of 10 and a tree number of 100 to analyze the cause of failure. The node records are compared with the historical data of equipment failures to identify the equipment overheating problem and classify the cause of failure as equipment failure. Next, the Apriori algorithm was used with a minimum support of 0.3 and a minimum confidence of 0.7 to extract rectification experience from historical cases and generate a suggestion: "Replace the transformer radiator and adjust the load distribution". The suggestion was stored in the MongoDB knowledge base and synchronized with the latest cases every week through incremental updates. Finally, the D3.js library was used to create an interactive flowchart to visualize the rectification suggestions into three key steps: checking the transformer temperature, replacing the radiator, and adjusting the load distribution. Each step was accompanied by detailed instructions and operation instructions, forming a complete closed loop of job evaluation and improvement.

[0051] Step S107: Based on the evaluation of the operation results, the data of the current review operation is compared and analyzed with the historical data of the same type of operation to identify abnormal fluctuations in operation efficiency and operation quality. If an abnormality is found, the person responsible for equipment management that caused the abnormality is traced back, and the operation plan is adjusted, the operation process nodes are optimized, and the data analysis of the on-site review operation nodes is completed.

[0052] Obtain the data of the current review operation and the historical operation data of the same type in the database, and use the ARIMA model to perform time series analysis on the operation efficiency and quality indicators based on the operation data to obtain abnormal fluctuation data. For abnormal fluctuation data, the K-means clustering algorithm is used to divide the abnormalities into three abnormal levels: slight, moderate and severe. The abnormal level is associated with the equipment ID in the equipment management database to determine the person responsible for the equipment management that caused the abnormality. According to the abnormal level, the genetic algorithm is used to optimize the operation plan. The chromosome encoding of the genetic algorithm includes operation time, personnel configuration and resource allocation. Process mining is performed on the operation process log in the operation plan to identify nodes with low efficiency or frequent quality problems. If nodes with low efficiency or frequent quality problems are identified, the optimized operation process nodes are redesigned. The optimized operation plan and operation process nodes are pushed to relevant personnel through the workflow management tool, and the execution status data is obtained using the real-time monitoring panel.

[0053] Specifically, the data of the current review operation and the historical data of the same type of operation are extracted from the database, where the same type of operation is defined based on the equipment type, operation nature and complexity. The ARIMA model is used to perform time series analysis on the operation efficiency and operation quality indicators, calculate the standard deviation and coefficient of variation, and identify abnormal fluctuations. According to the severity of the abnormal fluctuation, the K-means clustering algorithm is used to divide the abnormalities into three levels: slight, moderate and severe. Combined with the equipment ID and personnel management system in the equipment management database, the person responsible for the equipment management that caused the abnormality is traced, and whether the cause of the abnormality is related to equipment management is analyzed. For abnormalities of different levels, the genetic algorithm is used to optimize the operation plan. The chromosome encoding includes operation time, personnel configuration and resource allocation. New chromosomes are generated through operations with a crossover rate of 0.8 and a mutation rate of 0.1. The fitness function considers the weighted sum of operation efficiency and quality to generate a new operation plan. Based on process mining technology, the operation process log is analyzed to identify nodes with low efficiency or frequent quality problems, redesign the optimized operation process nodes, and verify the feasibility of the new process through discrete event simulation. The optimized operation plan and process nodes are pushed to relevant personnel through the workflow management tool, and the execution is tracked using the real-time monitoring panel to ensure that the optimization measures are implemented. In the process of distribution operation evaluation and optimization, the maintenance operation data of the 10kV transformer and the historical data of 100 similar operations in the past 6 months are first extracted from the Oracle database. The ARIMA (1,1,1) model is used to analyze the operation efficiency and quality indicators. The efficiency standard deviation is calculated to be 0.15, the quality coefficient of variation is 0.08, and 3 abnormal records are identified. The K-means clustering algorithm is used, where k=3 to classify the abnormalities, and 1 minor abnormality is obtained, whose efficiency is reduced by 10% and 2 medium abnormalities are obtained, whose quality is reduced by 15%. The device ID associated with the abnormality (such as TR-001, TR-002) is queried through the equipment management, and the responsible persons Zhang San and Li Si are matched. The genetic algorithm optimizes the operation plan, and the chromosome length is 20 bits, the first 10 bits represent the operation time (hours), and the last 10 bits represent the personnel and equipment configuration. The population size was set to 100, the iterations were 500 times, the crossover rate was 0.8, and the mutation rate was 0.1. The fitness function f = 0.6*efficiency + 0.4*quality, and the optimization plan was finally obtained: the operation time was shortened by 2 hours and 1 professional technician was added. Disco software was used for process mining, 1,000 operation logs were analyzed, and it was identified that the "equipment preheating" node took an average of 30 minutes, which was too long. The preheating step was redesigned to be parallelized, and the new process was verified by AnyLogic simulation software to reduce the total time by 15%. Finally, the optimized operation plan and process were pushed to 5 relevant personnel through the Jira workflow management tool, and the execution progress was displayed on the Grafana real-time monitoring panel, which was automatically updated every hour to ensure that the optimization measures were implemented.

[0054] Field operation data analysis system, including:

[0055] Information entry module, used to enter information on power distribution equipment type, installation location and planned operation time;

[0056] Equipment information collection module, used to collect equipment model, rated parameters, operating status and maintenance records;

[0057] The personnel information association module is used to obtain and associate the identity information and qualifications of the operating personnel with the operating tasks;

[0058] Environmental monitoring module, used to collect and monitor the environmental conditions of the operation nodes in real time to ensure safe operation;

[0059] The operation process analysis module is used to analyze the collected video images and determine whether the operation is compliant;

[0060] The work result evaluation module is used to compare and analyze work efficiency and quality, generate rectification suggestions and optimize.

[0061] The above description is only a preferred embodiment of the present application and an explanation of the technical principles used. Those skilled in the art should understand that the scope of the invention involved in the present application is not limited to the technical solution formed by a specific combination of the above technical features, but should also cover other technical solutions formed by any combination of the above technical features or their equivalent features without departing from the concept of the present application. For example, the above features are replaced with the technical features with similar functions disclosed in the present application (but not limited to) to form a technical solution.

Claims

1. A method for analyzing field operation data, characterized in that: The method comprises: According to the work order information of the power distribution field operation to be reviewed, the basic information of the power distribution equipment type, equipment installation location, and planned operation time of the operation task is entered into the unified database of operations to be reviewed; Through the electronic tags or QR codes of the power distribution equipment, the model specifications, rated parameters, operating status, and maintenance record equipment information of the equipment are collected to improve the equipment information in the database of the operation to be reviewed; According to the improved database of pending review operations, the identity information and qualifications of the operators involved in the pending review operations are obtained and associated with the corresponding operation tasks; Go to the corresponding work site according to the corresponding task, and collect and review the temperature, humidity, and noise environmental conditions of the work node environment in real time according to different work task types and work environment requirements. If the environmental indicators exceed the safety threshold of the work node task type, an early warning will be issued to the reviewer. If the work node environmental indicators meet the safety threshold, the review work process nodes will be collected in the form of video and image data; Analyze the collected video and image data of the review operation process nodes to determine whether the operators are operating in accordance with the operation node operating procedures, including whether they are using qualified operation tools and equipment and whether they comply with the operation safety protection requirements, and record the judgment results in the operation process node records; Based on the recorded operation process nodes, the completion efficiency of the operation tasks is compared and analyzed according to the planned completion time and the actual completion time of the operation tasks. The operation quality is scored in combination with the operation process node records to form an operation result evaluation. For the operation tasks that have not passed the review, the reasons for their failure are analyzed, and rectification suggestions are generated based on the equipment failure history and equipment environmental conditions; Based on the evaluation of the operation results, the data of this review operation will be compared and analyzed with the historical data of the same type of operation to identify abnormal fluctuations in operation efficiency and quality. If an abnormality is found, the person responsible for equipment management that caused the abnormality will be traced, and the operation plan will be adjusted, the operation process nodes will be optimized, and the data analysis of the on-site review operation nodes will be completed.

2. The method according to claim 1, wherein: According to the work order information of the power distribution field operation to be reviewed, the basic information of the power distribution equipment type, equipment installation location, and planned operation time of the operation task is entered into a unified database of operations to be reviewed, including: Acquire image information of the work order, extract text data from the image information, and identify key information including the type of power distribution equipment, the equipment installation location, and the planned operation time; Sending a similarity calculation instruction to a preset device type word library for the power distribution device type, wherein the preset device type word library calculates a similarity value using an edit distance algorithm according to the similarity calculation instruction; receiving a coordinate conversion instruction with device installation location information, wherein the coordinate conversion instruction is generated according to the device installation location information; It is determined whether the key information complies with the specified field mapping rule. If so, the key information is written into the database of jobs to be reviewed.

3. The method according to claim 1, wherein: The electronic tags or QR codes of the power distribution equipment are used to collect the model specifications, rated parameters, operating status, and maintenance record equipment information of the equipment, and to improve the equipment information of the database of the operation to be reviewed, including: Scan the electronic tag of the power distribution equipment with an RFID reader, or read the QR code of the equipment with an optical scanner to obtain a unique identifier and basic information of the equipment; Acquire a processable data format according to the unique identifier and basic information of the device; Querying device information matching the unique identifier of the device, the device information including model specifications, rated parameters, historical operating status and maintenance records; Comparing the processable data format with the historical records in the MySQL relational database, if the similarity is lower than a preset threshold, determining it as an inconsistent item; The device information is associated with corresponding records in the database of pending jobs, and the model specifications, rated parameters, operating status and maintenance records in the database of pending jobs are updated. If it is determined to be a new device or the device information does not exist in the database, a new device record is created in the database of pending jobs.

4. The method according to claim 1, wherein: The identity information and qualifications of the operators participating in the operation to be reviewed are obtained according to the improved database of the operation to be reviewed, and are associated with the corresponding operation tasks, including: Obtaining a list of personnel identification identifiers related to the job in a job database, wherein the list of personnel identification identifiers is extracted from the job database to be reviewed; According to the list of personnel identity identifiers, using a data integration tool to connect the database of pending operations and the human resources database, querying the identity information and qualification information of the operating personnel from the human resources database; Perform text analysis on the historical operation records of operators, extract keywords and construct professional ability vectors, and calculate the professional ability matching score of each person through cosine similarity; Calculate the number of times operators have completed similar operations and their success rates in the past six months to obtain historical operation experience evaluation results; The professional capability matching score and the historical work experience evaluation result are combined to generate a comprehensive score, and the comprehensive score is written back to the database of jobs to be reviewed to establish an association between personnel information and specific work tasks.

5. The method according to claim 1, wherein: According to the corresponding task, go to the corresponding work site, and collect and review the temperature, humidity, and noise environmental conditions of the work node environment in real time according to different work task types and work environment requirements; if the environmental indicators exceed the safety threshold of the work node task type, issue a warning to the reviewer; if the work node environmental indicators meet the safety threshold, the review work process node is collected in the form of video and image data, including: Obtaining the environmental safety threshold configuration from the preset MySQL database according to the job task type, wherein the environmental safety threshold configuration includes upper and lower limits of the three indicators of temperature, humidity and noise; Temperature and humidity sensors and noise sensors are used to arrange sensor nodes at the work site, and the sensor nodes collect environmental data once every preset time; Performing real-time analysis on the environmental data to determine whether the environmental data exceeds the environmental safety threshold; If the environmental data meets the environmental safety threshold for a preset period of time, the video and image data of the review operation process node are collected; Associating the environment data, the video and image data with the job task ID, and encapsulating the obtained associated data; It also includes: when collecting video and image data in the review operation process nodes, real-time detection and tracking of the motion trajectories of the node operators and equipment, and by analyzing the continuity and integrity of the motion trajectory, it is determined whether there is target loss or trajectory interruption due to occlusion or motion blur factors during the video and image data collection process, and the image data reliability assessment results are given.

6. The method according to claim 5, wherein: When collecting video and image data in the review process node, the motion trajectory of the node operator and equipment is detected and tracked in real time. By analyzing the continuity and integrity of the motion trajectory, it is determined whether there is target loss or trajectory interruption due to occlusion or motion blur during the video and image data collection process, and the image data reliability assessment results are given, including: Use target detection algorithms to process video frame sequences and obtain target bounding box coordinates of operators and equipment; According to the target bounding box coordinates, the detection result is smoothed to obtain a preliminary motion trajectory of the target; Calculating the Euclidean distance between adjacent frames in the target bounding box coordinate sequence, and determining that the trajectory is interrupted if the Euclidean distance is greater than a preset dynamic threshold; Identify the occluder in the video frame, calculate the overlapping area ratio between the target bounding box and the occluder, and determine that it is in an occlusion state if the overlapping area ratio exceeds a preset threshold; Calculating the pixel motion speed in the video frame sequence by using an optical flow method, and determining that the pixel motion speed exceeds a preset threshold value as motion blur; According to the determination results of the trajectory interruption, occlusion state and motion blur, the data reliability score is calculated by using the weighted average method.

7. The method according to claim 1, wherein: The collected video and image data of the review operation process nodes are analyzed to determine whether the operators operate in accordance with the operation node operating procedures, including whether qualified operation tools and equipment are used and whether the operation safety protection requirements are complied with, and the judgment results are recorded in the operation process node records, including: Perform image segmentation processing on the video frames to obtain target recognition results including operators, tools and working environment; According to the target recognition results, the position, shape and color feature information of each target is extracted; Comparing the characteristic information with the standard operation stored in the database to determine whether the operation behavior complies with the specification, including extracting the model and specification information of the operation tool in the video frame; According to the model and specification information, match it with the list of qualified tools required by the task to determine whether the tools used meet the requirements, including recording the tool's use angle and applied force data; Analyze the video frames using the OpenPose posture estimation algorithm to obtain body movement and position data of the operator; Processing the video frames using a deep learning model to identify the wearing status of safety protection equipment; Compare with the safety protection requirements of the operation to assess whether the operators strictly abide by the safety regulations; Performing time sequence analysis on the video frame sequence to obtain segmentation results of the operation process; For each time series node, a JSON format record is generated that includes operation behavior, tool usage, and security protection status; Compare the JSON format records with the standard operating procedures, calculate the operation compliance score, tool usage correctness and safety protection compliance rate, and record the calculation results in the operating process node records.

8. The method according to claim 1, wherein: The operation process nodes based on the records compare and analyze the completion efficiency of the operation tasks according to the planned completion time and the actual completion time of the operation tasks; score the operation quality in combination with the operation process node records to form an operation result evaluation; for the operation tasks that failed the review, analyze their failure reasons, and generate rectification suggestions based on the equipment failure history and equipment environmental conditions, including: Acquire a job flow node record from a database, wherein the job flow node record includes a planned completion time and an actual completion time of each node; Calculate the time difference percentage based on the planned completion time and the actual completion time, and use the ARIMA model to compare and analyze the completion efficiency of the task to obtain an efficiency score; If the efficiency score is lower than a preset threshold, the task efficiency is determined to be unsatisfactory. The weighted average method is used to calculate the operation quality score based on the operation standardization, tool usage correctness and safety protection compliance rate in the operation process node records; If the job quality score is lower than a preset threshold, the job task quality is determined to be unqualified; Using a random forest algorithm to analyze and determine the cause of failure of failed tasks, the random forest algorithm associates the operation process node records with the equipment failure history data to identify potential equipment problems and environmental factors; Based on the potential equipment problems and environmental factors, the causes of failure are classified into three categories: operation errors, equipment failures, and environmental impacts; Extract rectification experience from historical successful cases and generate targeted rectification suggestions based on the categories of failure causes.

9. The method according to claim 1, wherein: According to the evaluation of the operation results, the operation data of the current review is compared and analyzed with the historical operation data of the same type to identify abnormal fluctuations in operation efficiency and quality; If an abnormality is found, the person responsible for equipment management that caused the abnormality will be traced back, and the operation plan will be adjusted, the operation process nodes will be optimized, and the data analysis of the on-site review operation nodes will be completed, including: Obtain the data of the current audit operation and the historical data of the same type of operation in the database, and use the ARIMA model to perform time series analysis on the operation efficiency and quality indicators based on the operation data to obtain abnormal fluctuation data; For the abnormal fluctuation data, the K-means clustering algorithm is used to classify the abnormalities into three abnormality levels: slight, medium and severe. The abnormality level is associated with the device ID in the device management database to determine the person responsible for the device management that caused the abnormality; According to the abnormality level, optimizing the operation plan by using a genetic algorithm, wherein the chromosome encoding of the genetic algorithm includes operation time, staffing and resource allocation; Performing process mining on the job flow logs in the job plan to identify nodes with low efficiency or frequent quality problems; If nodes with low efficiency or frequent quality problems are identified, redesign the optimized operation process nodes; The optimized work plan and work process nodes are pushed to relevant personnel through the workflow management tool, and the execution status data is obtained using the real-time monitoring panel.

10. The field operation data analysis system is characterized by: The system comprises: Information entry module, used to enter information on power distribution equipment type, installation location and planned operation time; Equipment information collection module, used to collect equipment model, rated parameters, operating status and maintenance records; The personnel information association module is used to obtain and associate the identity information and qualifications of the operating personnel with the operating tasks; Environmental monitoring module, used to collect and monitor the environmental conditions of the operation nodes in real time to ensure safe operation; The operation process analysis module is used to analyze the collected video images and determine whether the operation is compliant; The work result evaluation module is used to compare and analyze work efficiency and quality, generate rectification suggestions and optimize.

Images