Resource trace management and control system based on power enterprise

Abstract

The invention discloses a resource trace management and control system based on a power enterprise, and relates to the technical field of resource management and control systems. The resource trace management and control system based on the power enterprise comprises an intelligent classification and integration unit, an image recognition and coding system, a dynamic back-end framework, a mobile terminal APP and an intelligent monitoring and early warning system. Through the intelligent classification and integration unit, the time cost of manual arrangement of technical requirements is greatly reduced, the accuracy is improved, and the timeliness and accuracy of the technical requirements are ensured. And secondly, the image identification and coding system utilizes a deep learning algorithm to realize efficient coding of standardized construction process pictures of the power transmission and transformation equipment, and coding efficiency and accuracy are further improved. And the dynamic back-end framework ensures high concurrent access and stable operation of the system through a micro-service architecture and a containerization technology, and meanwhile, the rapid deployment and elastic expansion capability of resources are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of resource management and control systems, and in particular to a resource traceability management and control system based on an electric power enterprise. Background Art

[0002] The amount of fixed assets in existing power companies accounts for a large proportion, generally exceeding 70% of the total assets. The number of fixed assets is huge and varied, including 16 categories such as transmission lines, substation equipment, and distribution lines, and they are widely distributed and difficult to manage. In addition, fixed assets are updated quickly, requiring an efficient management system to track and maintain them. In addition, power companies need to improve operational efficiency and reduce operating costs to cope with market competition and regulatory requirements. The resource management and control system can help companies achieve refined management, optimize resource allocation, and improve asset utilization. With the development of information technology, intelligence and automation have become important trends in the power industry. The resource management and control system integrates advanced information technology, control technology, and sensor technology to achieve intelligent management and automated control of the power system.

[0003] However, the existing resource management and control systems of power companies still have the following defects:

[0004] 1) Manual organization of technical requirements is inefficient and error-prone: Power companies need to manually organize technical requirements scattered in different regulations, specifications, and documents. This process is time-consuming, labor-intensive, and error-prone, making it difficult to ensure the timeliness and accuracy of technical requirements;

[0005] 2) Inefficient coding of equipment construction process images: Traditional coding methods rely on manual recognition and matching. For a large number of various standardized construction process images of power transmission and transformation equipment, the coding efficiency is low and the accuracy is not high, which makes it difficult to meet the needs of efficient management. Summary of the invention

[0006] In view of the deficiencies of the prior art, the present invention provides a resource traceability management and control system based on an electric power enterprise, which solves the problems of the prior art.

[0007] To achieve the above objectives, the present invention is implemented through the following technical solutions: a resource traceability management and control system based on power enterprises, comprising:

[0008] The intelligent classification and integration unit uses natural language processing and machine learning technologies to automatically classify the existing technical requirements scattered in different procedures, specifications and documents into different stages of the full-process technical supervision rules according to the release time. This step reduces the time cost of manual sorting and improves accuracy.

[0009] The image recognition and coding system is used for standardized construction process pictures of power transmission and transformation equipment. It uses image recognition technology to automatically extract key process features and match them with process names. Through deep learning algorithms, it ensures the unique correspondence between each picture and name, improving coding efficiency and accuracy.

[0010] Dynamic backend framework, through the adoption of microservice architecture, builds an extensible project backend framework, supports high concurrent access, and ensures stable operation of the system; at the same time, introduces containerization technology to quickly deploy and elastically scale resources;

[0011] The mobile APP integrates the standardized operation process of feasibility study review, and completes the management of review work from planning to archiving through GPS positioning and code scanning functions; and uses big data analysis to predict review progress and optimize resource allocation;

[0012] Intelligent monitoring and early warning system: Integrate system data, build an intelligent monitoring model, monitor equipment status and review progress in real time, trigger the early warning mechanism when an abnormality is identified, and improve emergency response speed.

[0013] Preferably, the intelligent classification and integration unit specifically includes:

[0014] The technical document AI classification module uses natural language processing (NLP) and machine learning algorithms to intelligently analyze technical documents and automatically identify key information in the documents, including equipment types and technical requirements. Based on the identification results, the document is automatically classified and summarized into the corresponding stages of the full-process technical supervision rules, and a timeline view is generated to show the evolution of technical requirements.

[0015] The dynamic update module regularly scans new technical documents, automatically classifies and adds them to the system to ensure the timeliness and completeness of technical requirements.

[0016] Preferably, the working steps of the technical document AI classification module specifically include:

[0017] a1 Preprocessing stage:

[0018] a1.1 Document format recognition: Add document format recognition parameters, such as PDF, Word, TXT, etc., to uniformly process documents of different formats;

[0019] a1.2 Language Identification: Add language identification parameters to support document processing in multiple languages ​​and ensure the wide applicability of the algorithm;

[0020] a1.3 OCR technology: For documents in non-text format (such as pictures), use OCR technology to convert them into processable text format;

[0021] a1.4 Text cleaning: remove irrelevant information in the document, such as advertisements, headers and footers, etc., to improve the accuracy of subsequent processing;

[0022] a2 Intelligent Analysis Stage:

[0023] a2.1 Keyword library: Build and continuously update keyword libraries such as equipment types and technical requirements to improve recognition accuracy;

[0024] a2.2 Contextual understanding: Add parameters based on contextual understanding to enable the algorithm to more accurately identify key information in documents;

[0025] a2.3 Deep Learning Model: Use deep learning algorithms (such as BERT, GPT, etc.) for natural language processing to improve the depth and breadth of text understanding;

[0026] a2.4 Multi-label classification: If a document contains multiple equipment types or technical requirements, a multi-label classification algorithm is used for processing;

[0027] a3 Classification and induction stage:

[0028] a3.1 Full-process technical supervision rules database: Build a full-process technical supervision rules database as a basis for classification and induction;

[0029] a3.2 Timestamp: Add timestamp parameters to each document to generate a timeline view;

[0030] a3.3 Rule-based classification: Combine supervised learning and rule engine to accurately classify documents;

[0031] a3.4 Timeline generation algorithm: Generate an intuitive timeline view based on timestamps and classification results to show the evolution of technical requirements.

[0032] Preferably, the working steps of the dynamic update module specifically include:

[0033] b1 New document scanning stage:

[0034] b1.1 Scanning frequency: Set the frequency parameters for regularly scanning new documents to ensure the timeliness of system updates;

[0035] b1.2 Source identification: Add document source identification parameters, such as internal systems, external websites, etc., so that documents from different sources can be processed uniformly;

[0036] b1.3 Incremental update algorithm: For newly added documents, use the incremental update algorithm to quickly process and classify them to avoid repeated processing of existing documents;

[0037] b2 Automatic classification and supplementation stage:

[0038] b2.1 Classification confidence: Set the classification confidence threshold parameter and conduct manual review or secondary classification for documents with low confidence;

[0039] b2.2 Online learning algorithm: Use online learning algorithm to classify new documents and continuously optimize model parameters to improve classification accuracy;

[0040] b2.3 System integration algorithm: Automatically add newly classified documents to the system to ensure the timeliness and completeness of technical requirements; at the same time, regularly check and update existing documents to ensure the accuracy and timeliness of system data.

[0041] Preferably, the image recognition and encoding system specifically includes:

[0042] The image coding module introduces deep learning models, such as convolutional neural networks (CNN), to perform high-precision recognition of standardized construction process images of power transmission and transformation equipment and automatically extract key features. Based on the recognition results, a unique code is generated for each image, and a one-to-one correspondence is established with the process name, which is stored in the database for quick retrieval.

[0043] The interactive image annotation tool allows users to manually correct the results of automatic annotation to improve the accuracy of encoding. The interactive image annotation tool provides a user feedback window to collect user feedback data as reference data for optimizing deep learning models.

[0044] Preferably, the working contents of the image encoding module specifically include:

[0045] c1. Image preprocessing:

[0046] Input image: Let the input image be I, whose size is H×W×C, where H is the height of the image, W is the width of the image, and C is the number of channels of the image (for RGB images, C=3);

[0047] Scaling: Scale the image to a fixed size H′×W′ to fit the input requirements of the CNN model;

[0048] Normalization: Normalize the pixel values ​​of the image to the interval [0, 1];

[0049] c2. Model structure of convolutional neural network (CNN) model:

[0050] Input layer: accepts the preprocessed image I′, with a size of H′×W′×C;

[0051] Convolutional layer: There are multiple convolutional layers, and the convolution operation of each layer is expressed as Conv(k, s, p), where k is the convolution kernel size, s is the step size, and p is the padding size; the convolution operation formula is: O = Conv(I′, K), where K is the convolution kernel and O is the output after convolution;

[0052] Pooling layer: There are multiple pooling layers, and the pooling operation of each layer is expressed as Pool(k, s), where k is the pooling kernel size and s is the step size; the pooling operation formula is: P = Pool(O), where P is the output after pooling;

[0053] Fully connected layer: There are multiple fully connected layers, each layer is represented as FC(n), where n is the number of neurons in the layer; the fully connected operation formula is: Z = FC(P), where Z is the output of the fully connected layer;

[0054] Output layer: Use the softmax function to output the classification probability, assuming the number of categories is N; the output layer formula is: Y = softmax (Z), where Y is the classification probability;

[0055] c3. Feature extraction and encoding generation:

[0056] Feature extraction: Extract the feature vector F from a certain layer of the CNN model (such as the output of the fully connected layer or the convolutional layer);

[0057] Eigenvector: Let the dimension of the eigenvector be D, then F∈RD;

[0058] Encoding generation: Using hash function h:R D →Z M Map the feature vector F to a unique integer code C, where M is the size of the hash space, and the encoding formula is expressed as: C = h(F);

[0059] c4. Database storage and retrieval:

[0060] Storage process: store the path of each image, the generated code and the corresponding process name in the database;

[0061] Retrieval process: According to the input code C, find the corresponding record in the database and return the process name or other relevant information.

[0062] Preferably, the microservice architecture adopts containerization technology (such as Docker) and Kubernetes cluster management technology to achieve rapid deployment, expansion and fault recovery of backend services, and is designed with a loosely coupled communication mechanism between services to ensure stable operation of the system in a high concurrency and low latency environment;

[0063] The dynamic backend framework also introduces an API gateway to uniformly manage the access rights and data flows of backend services, realize unified authentication, authorization and monitoring of services; and establish a data governance framework to standardize data standards, quality and security, and ensure the accuracy and consistency of system data.

[0064] Preferably, the mobile APP is built based on the cross-platform framework of React Native or Flutter, supports one-time writing and multi-terminal operation, reduces development costs and maintenance difficulties; and is equipped with AI technologies, such as natural language understanding and knowledge graphs, to provide reviewers with intelligent suggestions, risk warnings and decision support.

[0065] Preferably, the intelligent monitoring and early warning system includes:

[0066] Real-time data analysis and prediction: Use big data analysis and machine learning algorithms to analyze system data in real time, predict equipment failures, review key indicators of progress, and design visual reports and dashboards to display analysis results and forecast trends;

[0067] Intelligent early warning and emergency response: Establish an intelligent early warning mechanism to automatically trigger early warning notifications based on analysis and prediction results, reminding relevant personnel to take countermeasures; there is also an emergency response process to quickly initiate emergency plans and reduce losses in emergency situations such as equipment failure or review delays.

[0068] Preferably, the working contents of the intelligent monitoring and early warning system specifically include:

[0069] d1 Big Data Analysis and Machine Learning Algorithms:

[0070] Algorithm selection: Select a suitable machine learning algorithm based on the data type and prediction target, which can be one of linear regression, logistic regression, support vector machine, or neural network;

[0071] Feature engineering: extract useful features from raw data and construct feature vector X;

[0072] Model training: Use historical data D = {(X1, y1), (X2, y2), ..., (X m ,y m )} training model, where X i is the eigenvector, y i is the target value (such as equipment failure rate, review progress completion rate, etc.);

[0073] Real-time analysis: Use the trained model to analyze real-time data and predict future equipment failure rates y^ fault and review progress completion rate y^ progress ;

[0074] Visual reports and dashboards: Design visual reports to display analysis results and forecast trends, including equipment failure rate trend charts and review progress completion rate bar charts; design real-time updated dashboards to display key indicators, including equipment failure warning levels and review progress delay risks;

[0075] d2 Intelligent Early Warning and Emergency Response:

[0076] Intelligent early warning mechanism:

[0077] Warning threshold setting: According to historical data and business needs, set the warning threshold T of equipment failure rate and review progress completion rate fault and T progress ;

[0078] Warning triggering condition: When the predicted equipment failure rate y^ fault >T fault Or the predicted review progress completion rate y^ progress <T progress When the warning notification is triggered;

[0079] Warning notification method: remind relevant personnel to take countermeasures through SMS, email, APP push, etc.

[0080] d3 Emergency Response Process:

[0081] Emergency plan formulation: formulate emergency plans for emergency situations such as equipment failure and review delays, and clarify emergency measures and responsible persons;

[0082] Emergency response initiation: When an early warning is triggered, the emergency plan is quickly initiated and relevant personnel are organized to handle the emergency;

[0083] Loss assessment and reduction: Monitor and evaluate the emergency response process, adjust emergency measures in a timely manner, and reduce losses.

[0084] The present invention provides a resource traceability management and control system based on power enterprises. Compared with the prior art, it has the following beneficial effects:

[0085] 1. The resource traceability management and control system based on power enterprises has greatly reduced the time cost of manually sorting technical requirements through intelligent classification and integration units, while improving accuracy and ensuring the timeliness and accuracy of technical requirements. Secondly, the image recognition and coding system uses deep learning algorithms to achieve efficient coding of standardized construction process pictures of power transmission and transformation equipment, further improving coding efficiency and accuracy. The dynamic backend framework ensures high concurrent access and stable operation of the system through microservice architecture and containerization technology, while improving the rapid deployment and elastic scaling capabilities of resources. The introduction of mobile APP not only reduces development costs and maintenance difficulties, but also provides intelligent suggestions and risk warnings to reviewers through AI technology, optimizing resource allocation. Finally, the intelligent monitoring and early warning system can monitor equipment status and review progress in real time, trigger the early warning mechanism in time, and significantly improve the emergency response speed. These improvements have jointly promoted the intelligence and efficiency of resource management and control of power enterprises.

[0086] 2. The resource traceability management and control system based on power enterprises realizes intelligent analysis, automatic classification and summarization of technical documents through the technical document AI classification module, significantly improving the efficiency and accuracy of document processing. This module uses natural language processing and deep learning algorithms to accurately identify key information in documents and classify them into the corresponding stages of the technical supervision details of the whole process. At the same time, it generates an intuitive timeline view to show the evolution of technical requirements. Secondly, the dynamic update module ensures that the system can regularly scan and update new technical documents, maintaining the real-time and integrity of technical requirements. This module uses incremental update algorithms and online learning algorithms to achieve rapid processing and classification of new documents, and continuously optimizes model parameters to improve classification accuracy. These improvements have jointly improved the efficiency and quality of power companies' management of technical documents, and provided strong support for the supervision and implementation of technical requirements. At the same time, it also reduces the cost of manual sorting and maintenance, and promotes the intelligence and efficiency of resource management and control of power companies.

[0087] 3. This resource traceability management and control system based on power companies, by introducing deep learning models, such as convolutional neural networks (CNN), the system achieves high-precision recognition of standardized construction process pictures of power transmission and transformation equipment, can automatically extract key features and generate unique codes, greatly improving the efficiency and accuracy of coding. Secondly, the interactive picture annotation tool allows users to manually correct the results of automatic annotation, further improving the accuracy of coding, and optimizes the deep learning model through user feedback data, forming a closed loop of continuous improvement. In addition, the system also has efficient database storage and retrieval functions, which can quickly locate and return relevant information such as process names, providing strong support for the construction process management of power companies. These improvements have jointly promoted the application of image recognition technology in power companies and improved the level of intelligent construction process management.

[0088] 4. This resource traceability management and control system based on power companies can predict equipment failures and review progress in real time through big data analysis and machine learning algorithms, provide a scientific basis for decision-making, and intuitively display analysis results through visual reports and dashboards, enhancing the transparency and efficiency of management. The establishment of an intelligent early warning mechanism enables early warning notifications to be automatically triggered, promptly reminding relevant personnel to take countermeasures, and effectively avoiding potential risks. In addition, the improvement of the emergency response process ensures that emergency plans can be quickly activated in emergency situations, reducing losses and improving the company's emergency response capabilities. These improvements have jointly improved the operational management level of power companies and enhanced their competitiveness and sustainable development capabilities. BRIEF DESCRIPTION OF THE DRAWINGS

[0089] Figure 1 It is a system principle block diagram of the present invention;

[0090] Figure 2 It is a schematic diagram of the main process framework of the present invention. DETAILED DESCRIPTION

[0091] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not 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.

[0092] See also Figure 1-Figure 2 , the present invention provides the following four technical solutions:

[0093] The first implementation method: a resource traceability management and control system based on an electric power enterprise, comprising:

[0094] The intelligent classification and integration unit uses natural language processing and machine learning technologies to automatically classify the existing technical requirements scattered in different procedures, specifications and documents into different stages of the full-process technical supervision rules according to the release time. This step reduces the time cost of manual sorting and improves accuracy.

[0095] The image recognition and coding system is used for standardized construction process pictures of power transmission and transformation equipment. It uses image recognition technology to automatically extract key process features and match them with process names. Through deep learning algorithms, it ensures the unique correspondence between each picture and name, improving coding efficiency and accuracy.

[0096] The dynamic backend framework uses a microservice architecture to build an extensible project backend framework that supports high-concurrency access and ensures stable system operation. At the same time, containerization technology is introduced to quickly deploy and elastically scale resources. The microservice architecture uses containerization technology (such as Docker) and Kubernetes cluster management technology to achieve rapid deployment, expansion and fault recovery of backend services, and is designed with a loosely coupled communication mechanism between services to ensure stable operation of the system in a high-concurrency, low-latency environment. The dynamic backend framework also introduces an API gateway to uniformly manage the access rights and data flows of backend services, and achieve unified authentication, authorization and monitoring of services. A data governance framework is established to standardize data standards, quality and security to ensure the accuracy and consistency of system data.

[0097] The mobile APP integrates the standardized operation process of feasibility study review, and completes the management of review work from planning to archiving through GPS positioning and code scanning functions. It also uses big data analysis to predict the progress of the review and optimize resource allocation. The mobile APP is built based on the cross-platform framework of React Native or Flutter, supports one-time writing and multi-terminal operation, reducing development costs and maintenance difficulties. It is also equipped with AI technologies such as natural language understanding and knowledge graphs to provide reviewers with intelligent suggestions, risk warnings and decision-making support.

[0098] Intelligent monitoring and early warning system: Integrate system data, build an intelligent monitoring model, monitor equipment status and review progress in real time, trigger the early warning mechanism when an abnormality is identified, and improve emergency response speed.

[0099] Through intelligent classification and integration units, the time cost of manually sorting technical requirements is greatly reduced, while the accuracy is improved, ensuring the timeliness and accuracy of technical requirements. Secondly, the image recognition and coding system uses deep learning algorithms to achieve efficient coding of standardized construction process pictures of power transmission and transformation equipment, further improving coding efficiency and accuracy. The dynamic backend framework ensures high concurrent access and stable operation of the system through microservice architecture and containerization technology, while improving the rapid deployment and elastic scaling capabilities of resources. The introduction of mobile APP not only reduces development costs and maintenance difficulties, but also provides intelligent suggestions and risk warnings to reviewers through AI technology, optimizing resource allocation. Finally, the intelligent monitoring and early warning system can monitor equipment status and review progress in real time, trigger the early warning mechanism in time, and significantly improve the emergency response speed. These improvements have jointly promoted the intelligence and efficiency of resource management and control of power companies.

[0100] The second implementation mode is mainly different from the first implementation mode in that the intelligent classification and integration unit specifically includes:

[0101] The technical document AI classification module uses natural language processing (NLP) and machine learning algorithms to intelligently analyze technical documents and automatically identify key information in the documents, including equipment types and technical requirements. Based on the identification results, the document is automatically classified and summarized into the corresponding stages of the full-process technical supervision rules, and a timeline view is generated to show the evolution of technical requirements.

[0102] The dynamic update module regularly scans new technical documents, automatically classifies and adds them to the system to ensure the timeliness and completeness of technical requirements.

[0103] The working steps of the technical document AI classification module include:

[0104] a1 Preprocessing stage:

[0105] a1.1 Document format recognition: Add document format recognition parameters, such as PDF, Word, TXT, etc., to uniformly process documents of different formats;

[0106] a1.2 Language Identification: Add language identification parameters to support document processing in multiple languages ​​and ensure the wide applicability of the algorithm;

[0107] a1.3 OCR technology: For documents in non-text format (such as pictures), use OCR technology to convert them into processable text format;

[0108] a1.4 Text cleaning: remove irrelevant information in the document, such as advertisements, headers and footers, etc., to improve the accuracy of subsequent processing;

[0109] a2 Intelligent Analysis Stage:

[0110] a2.1 Keyword library: Build and continuously update keyword libraries such as equipment types and technical requirements to improve recognition accuracy;

[0111] a2.2 Contextual understanding: Add parameters based on contextual understanding to enable the algorithm to more accurately identify key information in documents;

[0112] a2.3 Deep Learning Model: Use deep learning algorithms (such as BERT, GPT, etc.) for natural language processing to improve the depth and breadth of text understanding;

[0113] a2.4 Multi-label classification: If a document contains multiple equipment types or technical requirements, a multi-label classification algorithm is used for processing;

[0114] a3 Classification and induction stage:

[0115] a3.1 Full-process technical supervision rules database: Build a full-process technical supervision rules database as a basis for classification and induction;

[0116] a3.2 Timestamp: Add timestamp parameters to each document to generate a timeline view;

[0117] a3.3 Rule-based classification: Combine supervised learning and rule engine to accurately classify documents;

[0118] a3.4 Timeline generation algorithm: Generate an intuitive timeline view based on timestamps and classification results to show the evolution of technical requirements.

[0119] The technical document AI classification module enables intelligent analysis, automatic classification and summarization of technical documents, significantly improving the efficiency and accuracy of document processing. The module uses natural language processing and deep learning algorithms to accurately identify key information in documents and classify them into the corresponding stages of the full-process technical supervision details, while generating an intuitive timeline view that shows the evolution of technical requirements.

[0120] The working steps of the dynamic update module include:

[0121] b1 New document scanning stage:

[0122] b1.1 Scanning frequency: Set the frequency parameters for regularly scanning new documents to ensure the timeliness of system updates;

[0123] b1.2 Source identification: Add document source identification parameters, such as internal systems, external websites, etc., so that documents from different sources can be processed uniformly;

[0124] b1.3 Incremental update algorithm: For newly added documents, use the incremental update algorithm to quickly process and classify them to avoid repeated processing of existing documents;

[0125] b2 Automatic classification and supplementation stage:

[0126] b2.1 Classification confidence: Set the classification confidence threshold parameter and conduct manual review or secondary classification for documents with low confidence;

[0127] b2.2 Online learning algorithm: Use online learning algorithm to classify new documents and continuously optimize model parameters to improve classification accuracy;

[0128] b2.3 System integration algorithm: Automatically add newly classified documents to the system to ensure the timeliness and completeness of technical requirements; at the same time, regularly check and update existing documents to ensure the accuracy and timeliness of system data.

[0129] The dynamic update module ensures that the system can regularly scan and update new technical documents, maintaining the real-time and integrity of technical requirements. The module uses incremental update algorithms and online learning algorithms to achieve rapid processing and classification of new documents, and continuously optimizes model parameters to improve classification accuracy.

[0130] These improvements have jointly improved the efficiency and quality of power companies' management of technical documents, provided strong support for the supervision and implementation of technical requirements, reduced the cost of manual sorting and maintenance, and promoted the intelligent and efficient resource management and control of power companies.

[0131] The third implementation mode is mainly different from the first implementation mode in that the image recognition and encoding system specifically includes:

[0132] The image coding module introduces deep learning models, such as convolutional neural networks (CNN), to perform high-precision recognition of standardized construction process images of power transmission and transformation equipment and automatically extract key features. Based on the recognition results, a unique code is generated for each image, and a one-to-one correspondence is established with the process name, which is stored in the database for quick retrieval.

[0133] The interactive image annotation tool allows users to manually correct the results of automatic annotation to improve the accuracy of encoding. The interactive image annotation tool provides a user feedback window to collect user feedback data as reference data for optimizing deep learning models.

[0134] The work content of the image coding module specifically includes:

[0135] c1. Image preprocessing:

[0136] Input image: Let the input image be I, whose size is H×W×C, where H is the height of the image, W is the width of the image, and C is the number of channels of the image (for RGB images, C=3);

[0137] Scaling: Scale the image to a fixed size H′×W′ to fit the input requirements of the CNN model;

[0138] Normalization: Normalize the pixel values ​​of the image to the interval [0, 1];

[0139] c2. Model structure of convolutional neural network (CNN) model:

[0140] Input layer: accepts the preprocessed image I′, with a size of H′×W′×C;

[0141] Convolutional layer: There are multiple convolutional layers, and the convolution operation of each layer is expressed as Conv(k, s, p), where k is the convolution kernel size, s is the step size, and p is the padding size; the convolution operation formula is: O = Conv(I′, K), where K is the convolution kernel and O is the output after convolution;

[0142] Pooling layer: There are multiple pooling layers, and the pooling operation of each layer is expressed as Pool(k, s), where k is the pooling kernel size and s is the step size; the pooling operation formula is: P = Pool(O), where P is the output after pooling;

[0143] Fully connected layer: There are multiple fully connected layers, each layer is represented as FC(n), where n is the number of neurons in the layer; the fully connected operation formula is: Z = FC(P), where Z is the output of the fully connected layer;

[0144] Output layer: Use the softmax function to output the classification probability, assuming the number of categories is N; the output layer formula is: Y = softmax (Z), where Y is the classification probability;

[0145] c3. Feature extraction and encoding generation:

[0146] Feature extraction: Extract the feature vector F from a certain layer of the CNN model (such as the output of the fully connected layer or the convolutional layer);

[0147] Eigenvector: Let the dimension of the eigenvector be D, then F∈RD;

[0148] Encoding generation: Using hash function h:R D →Z M Map the feature vector F to a unique integer code C, where M is the size of the hash space, and the encoding formula is expressed as: C = h(F);

[0149] c4. Database storage and retrieval:

[0150] Storage process: store the path of each image, the generated code and the corresponding process name in the database;

[0151] Retrieval process: According to the input code C, find the corresponding record in the database and return the process name or other relevant information.

[0152] By introducing deep learning models, such as convolutional neural networks (CNN), the system has achieved high-precision recognition of standardized construction process images of power transmission and transformation equipment, and can automatically extract key features and generate unique codes, greatly improving the efficiency and accuracy of coding. Secondly, the interactive image annotation tool allows users to manually correct the results of automatic annotation, further improving the accuracy of coding, and optimizes the deep learning model through user feedback data, forming a closed loop of continuous improvement. In addition, the system also has efficient database storage and retrieval functions, which can quickly locate and return relevant information such as process names, providing strong support for the construction process management of power companies. These improvements have jointly promoted the application of image recognition technology in power companies and improved the level of intelligent construction process management.

[0153] The fourth implementation mode is mainly different from the first implementation mode in that the intelligent monitoring and early warning system includes:

[0154] Real-time data analysis and prediction: Use big data analysis and machine learning algorithms to analyze system data in real time, predict equipment failures, review key indicators of progress, and design visual reports and dashboards to display analysis results and forecast trends;

[0155] Intelligent early warning and emergency response: Establish an intelligent early warning mechanism to automatically trigger early warning notifications based on analysis and prediction results, reminding relevant personnel to take countermeasures; there is also an emergency response process to quickly initiate emergency plans and reduce losses in emergency situations such as equipment failure or review delays.

[0156] The work content of the intelligent monitoring and early warning system includes:

[0157] d1 Big Data Analysis and Machine Learning Algorithms:

[0158] Algorithm selection: Select a suitable machine learning algorithm based on the data type and prediction target, which can be one of linear regression, logistic regression, support vector machine, or neural network;

[0159] Feature engineering: extract useful features from raw data and construct feature vector X;

[0160] Model training: Use historical data D = {(X1, y1), (X2, y2), ..., (X m ,y m )} training model, where X i is the eigenvector, y i is the target value (such as equipment failure rate, review progress completion rate, etc.);

[0161] Real-time analysis: Use the trained model to analyze real-time data and predict future equipment failure rates y^ fault and review progress completion rate y^progress ;

[0162] Visual reports and dashboards: Design visual reports to display analysis results and forecast trends, including equipment failure rate trend charts and review progress completion rate bar charts; design real-time updated dashboards to display key indicators, including equipment failure warning levels and review progress delay risks;

[0163] d2 Intelligent Early Warning and Emergency Response:

[0164] Intelligent early warning mechanism:

[0165] Warning threshold setting: According to historical data and business needs, set the warning threshold T of equipment failure rate and review progress completion rate fault and T progress ;

[0166] Warning triggering condition: When the predicted equipment failure rate y^ fault >T fault Or the predicted review progress completion rate y^ progress <T progress When the warning notification is triggered;

[0167] Warning notification method: remind relevant personnel to take countermeasures through SMS, email, APP push, etc.

[0168] d3 Emergency Response Process:

[0169] Emergency plan formulation: formulate emergency plans for emergency situations such as equipment failure and review delays, and clarify emergency measures and responsible persons;

[0170] Emergency response initiation: When an early warning is triggered, the emergency plan is quickly initiated and relevant personnel are organized to handle the emergency;

[0171] Loss assessment and reduction: Monitor and evaluate the emergency response process, adjust emergency measures in a timely manner, and reduce losses.

[0172] Through big data analysis and machine learning algorithms, the system can predict equipment failures and review progress in real time, provide a scientific basis for decision-making, and intuitively display analysis results through visual reports and dashboards, enhancing the transparency and efficiency of management. The establishment of an intelligent early warning mechanism enables early warning notifications to be automatically triggered, promptly reminding relevant personnel to take countermeasures, and effectively avoiding potential risks. In addition, the improvement of the emergency response process ensures that emergency plans can be quickly activated in emergency situations, reducing losses and improving the company's emergency response capabilities. These improvements have jointly improved the operational management level of power companies and enhanced their competitiveness and sustainable development capabilities.

[0173] Meanwhile, the contents not described in detail in this specification belong to the prior art known to those skilled in the art.

[0174] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device.

[0175] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A resource traceability management and control system based on power enterprises, characterized in that: include: The intelligent classification and integration unit uses natural language processing and machine learning technologies to automatically classify the existing technical requirements scattered in different procedures, specifications and documents into different stages of the full-process technical supervision details according to the release time according to the equipment classification; The image recognition and coding system is used for standardized construction process pictures of power transmission and transformation equipment. It uses image recognition technology to automatically extract key process features and match the codes with process names. Through deep learning algorithms, it ensures the unique correspondence between each picture and the name. Dynamic backend framework, through the adoption of microservice architecture, builds an extensible project backend framework that supports high concurrent access; at the same time, introduces containerization technology to quickly deploy and elastically scale resources; The mobile APP integrates the standardized operation process of feasibility study review and completes the management of review work from planning to filing through GPS positioning and code scanning functions; And use big data analysis to predict review progress and optimize resource allocation; Intelligent monitoring and early warning system: Integrate system data, build an intelligent monitoring model, monitor equipment status and review progress in real time, and trigger an early warning mechanism when an abnormality is identified.

2. According to claim 1, a resource traceability management and control system based on an electric power enterprise is characterized in that: The intelligent classification and integration unit specifically includes: The technical document AI classification module uses natural language processing and machine learning algorithms to intelligently analyze technical documents and automatically identify key information in the documents, including equipment types and technical requirements. Based on the identification results, the module automatically classifies the documents and summarizes them into the corresponding stages of the full-process technical supervision rules. It also generates a timeline view to show the evolution of technical requirements. Dynamic update module, regularly scans new technical documents, and automatically categorizes and adds them to the system.

3. According to claim 2, a resource traceability management and control system based on an electric power enterprise is characterized in that: The working steps of the technical document AI classification module include: a1 Preprocessing stage: a1.1 Document format recognition: Add document format recognition parameters to uniformly process documents of different formats; a1.2 Language Identification: Add language identification parameters to support document processing in multiple languages; a1.3 OCR technology: For documents in non-text format, use OCR technology to convert them into processable text format; a1.4 Text cleaning: remove irrelevant information from the document; a2 Intelligent Analysis Stage: a2.1 Keyword library: Build and continuously update the keyword library of equipment types and technical requirements to improve recognition accuracy; a2.2 Contextual understanding: Add parameters based on contextual understanding to enable the algorithm to more accurately identify key information in documents; a2.3 Deep Learning Model: Use deep learning algorithms for natural language processing to improve the depth and breadth of text understanding; a2.4 Multi-label classification: If a document contains multiple equipment types or technical requirements, a multi-label classification algorithm is used for processing; a3 Classification and induction stage: a3.1 Full-process technical supervision rules database: Build a full-process technical supervision rules database as a basis for classification and induction; a3.2 Timestamp: Add timestamp parameters to each document to generate a timeline view; a3.3 Rule-based classification: Combine supervised learning and rule engine to accurately classify documents; a3.4 Timeline generation algorithm: Generate an intuitive timeline view based on timestamps and classification results to show the evolution of technical requirements.

4. According to claim 2, a resource traceability management and control system based on an electric power enterprise is characterized in that: The working steps of the dynamic update module include: b1 New document scanning stage: b1.1 Scanning frequency: Set the frequency parameters for regularly scanning new documents to ensure the timeliness of system updates; b1.2 Source Identification: Add document source identification parameters to enable unified processing of documents from different sources; b1.3 Incremental update algorithm: For newly added documents, use the incremental update algorithm to quickly process and classify them to avoid repeated processing of existing documents; b2 Automatic classification and supplementation stage: b2.1 Classification confidence: Set the classification confidence threshold parameter and conduct manual review or secondary classification for documents with low confidence; b2.2 Online learning algorithm: Use online learning algorithm to classify new documents and continuously optimize model parameters to improve classification accuracy; b2.3 System integration algorithm: Automatically add newly classified documents to the system to ensure the timeliness and completeness of technical requirements; at the same time, regularly check and update existing documents to ensure the accuracy and timeliness of system data.

5. According to claim 1, a resource traceability management and control system based on an electric power enterprise is characterized in that: The image recognition and encoding system specifically includes: The image coding module introduces a deep learning model to perform high-precision recognition of standardized construction process images of power transmission and transformation equipment and automatically extract key features. Based on the recognition results, a unique code is generated for each image, and a one-to-one correspondence is established with the process name and stored in the database. The interactive image annotation tool allows users to manually correct the results of automatic annotation. The interactive image annotation tool provides a user feedback window to collect user feedback data as reference data for optimizing deep learning models.

6. The resource traceability management and control system based on power enterprises according to claim 1 is characterized by: The working contents of the image encoding module specifically include: c1. Image preprocessing: Input image: Let the input image be I, and its size is H×W×C, where H is the height of the image, W is the width of the image, and C is the number of channels of the image; Scaling: Scale the image to a fixed size H′×W′ to fit the input requirements of the CNN model; Normalization: Normalize the pixel values ​​of the image to the interval [0, 1]; c2. Model structure of convolutional neural network (CNN) model: Input layer: accepts the preprocessed image I′, with a size of H′×W′×C; Convolutional layer: There are multiple convolutional layers, and the convolution operation of each layer is expressed as Conv(k, s, p), where k is the convolution kernel size, s is the step size, and p is the padding size; the convolution operation formula is: O = Conv(I′, K), where K is the convolution kernel and O is the output after convolution; Pooling layer: There are multiple pooling layers, and the pooling operation of each layer is expressed as Pool(k, s), where k is the pooling kernel size and s is the step size; the pooling operation formula is: P = Pool(O), where P is the output after pooling; Fully connected layer: There are multiple fully connected layers, each layer is represented as FC(n), where n is the number of neurons in the layer; the fully connected operation formula is: Z = FC(P), where Z is the output of the fully connected layer; Output layer: Use the softmax function to output the classification probability, assuming the number of categories is N; the output layer formula is: Y = softmax (Z), where Y is the classification probability; c3. Feature extraction and encoding generation: Feature extraction: extract the feature vector F from a certain layer of the CNN model; Eigenvector: Let the dimension of the eigenvector be D, then F∈RD; Encoding generation: Using hash function h:R D →Z M Map the feature vector F to a unique integer code C, where M is the size of the hash space, and the encoding formula is expressed as: C = h(F); c4. Database storage and retrieval: Storage process: store the path of each image, the generated code and the corresponding process name in the database; Retrieval process: According to the input code C, find the corresponding record in the database and return the process name or other relevant information.

7. The resource traceability management and control system based on power enterprises according to claim 1 is characterized by: The microservice architecture adopts containerization technology and Kubernetes cluster management technology, and is designed with a loosely coupled communication mechanism between services; The dynamic backend framework also introduces an API gateway to uniformly manage access rights and data flows of backend services; and establishes a data governance framework to regulate data standards, quality and security.

8. The resource traceability management and control system based on power enterprises according to claim 1 is characterized by: The mobile APP is built based on the cross-platform framework of React Native or Flutter, supports one-time writing and multi-terminal operation; and is equipped with AI technology to provide reviewers with intelligent suggestions, risk warnings and decision-making support.

9. The resource traceability management and control system based on power enterprises according to claim 1 is characterized by: The intelligent monitoring and early warning system includes: Real-time data analysis and prediction: Use big data analysis and machine learning algorithms to analyze system data in real time, predict equipment failures, review key indicators of progress, and design visual reports and dashboards to display analysis results and forecast trends; Intelligent early warning and emergency response: Establish an intelligent early warning mechanism to automatically trigger early warning notifications based on analysis and prediction results; also have an emergency response process to quickly initiate emergency plans in the event of equipment failure or review delays.

10. A resource traceability management and control system based on electric power enterprises according to claim 9, characterized in that: The work contents of the intelligent monitoring and early warning system specifically include: d1 Big Data Analysis and Machine Learning Algorithms: Algorithm selection: Select a suitable machine learning algorithm based on the data type and prediction target, which can be one of linear regression, logistic regression, support vector machine, or neural network; Feature engineering: extract useful features from raw data and construct feature vector X; Model training: Use historical data D = {(X1, y1), (X2, y2), ..., (X m ,y m )} training model, where X i is the eigenvector, y i is the target value; Real-time analysis: Use the trained model to analyze real-time data and predict future equipment failure rates y^ fault and review progress completion rate y^ progress ; Visual reports and dashboards: Design visual reports to display analysis results and forecast trends, including equipment failure rate trend charts and review progress completion rate bar charts; design real-time updated dashboards to display key indicators, including equipment failure warning levels and review progress delay risks; d2 Intelligent Early Warning and Emergency Response: Intelligent early warning mechanism: Warning threshold setting: According to historical data and business needs, set the warning threshold T of equipment failure rate and review progress completion rate fault and T progress ; Warning triggering condition: When the predicted equipment failure rate y^ fault >T fault Or the predicted review progress completion rate y^ progress <T progress When the alarm is triggered, an early warning notification is triggered; Warning notification method: through SMS, email, APP push, remind relevant personnel to take countermeasures; d3 Emergency Response Process: Emergency plan formulation: formulate emergency plans for emergency situations such as equipment failure and review delays, and clarify emergency measures and responsible persons; Emergency response initiation: When an early warning is triggered, the emergency plan is quickly initiated and relevant personnel are organized to handle the emergency; Loss assessment and reduction: Monitor and evaluate the emergency response process, adjust emergency measures in a timely manner, and reduce losses.

Images