AI system and processing method for power grid business expansion business

Abstract

The invention relates to an AI system for power grid business expansion business and a processing method. The AI system comprises an AI acceptance sample model database, an iterative training module, a business system module and a data interface module. The AI acceptance sample model database comprises a qualified sample library model set and a defect sample library model set; the iterative training module provides iterative training for the system through OPEN AI, a multi-scale feature extraction network and other algorithms based on a model set in the AI acceptance model database, establishes a mapping relation, processes submitted acceptance data according to the established mapping relation, and determines whether the power equipment is qualified or not; the business system module provides an input port for acceptance data; the data interface module carries out standardization processing on the input acceptance data and outputs the data to the iteration smooth module; and the iterative training module calculates a result online and outputs the result to the service system module through the data interface module so as to store and display whether the acceptance result is qualified or not. According to the invention, the business expansion acceptance efficiency is improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of power grid expansion and installation, and specifically relates to an AI system and processing method for power grid expansion and installation services. Background Art

[0002] In the business expansion application process, the completion inspection is a key link to control the safe access of customer equipment to the grid, which is directly related to the safe and stable operation of the power grid and the continuous and reliable power supply of customers. The account manager needs to conduct a comprehensive on-site acceptance of the equipment, facilities and technical data after completion, and can only apply for power supply after the acceptance is qualified.

[0003] In the traditional model, the acceptance process mainly relies on the account manager to go back and forth to the customer site for acceptance. The acceptance process is "acceptance-rectification-acceptance". Only when all aspects pass the acceptance can the next stage be entered. In this process, due to the lack of experience of the customer, the problem cannot be rectified in one go, which leads to multiple rectifications and multiple round trips, resulting in problems such as prolonged rectification cycle and low efficiency. Summary of the invention

[0004] In view of the deficiencies in the prior art, the present invention proposes an AI system and processing method for power grid expansion and installation services, which can reduce processing time and improve work efficiency.

[0005] One of the above objects of the present invention is achieved by the following technical solution:

[0006] An AI system for power grid expansion and installation services, characterized by: comprising an AI acceptance sample model database, an iterative training module, a business system module and a data interface module;

[0007] The AI ​​acceptance sample model database includes multiple model sets such as a qualified sample library and a defective sample library. Each model set stores multiple AI samples of the same type. The AI ​​acceptance samples are annotated with knowledge by senior experts. The AI ​​acceptance sample model database is used to provide standard data for iterative training for the iterative training module.

[0008] The iterative training module provides iterative training for the system based on the model set in the AI ​​acceptance model database through OPEN AI, multi-scale feature extraction network, class activation mapping and other algorithms, establishes a mapping relationship, and processes the acceptance image data submitted by the user according to the established mapping relationship to determine whether the submitted acceptance data is qualified;

[0009] The business system module is used to provide an input port for acceptance document data;

[0010] The data interface module complies with the power grid network security strategy. The data interface module standardizes the input acceptance image data, pre-processes the image content and image quality, and outputs them to the iterative training module; the iterative training module calculates the results online and outputs them to the business system module through the data interface module to realize the storage and display of whether the acceptance results are qualified.

[0011] Moreover, the business system module includes a marketing business system and an online State Grid mobile power management system, and the two can achieve data synchronization through the data middle platform.

[0012] Moreover, the data interface module includes the "E-Assistant for Electricity Management" module of the online State Grid mobile electricity management system, and an image data upload interface is added to the module. Users can upload the installation status of on-site electrical equipment through the interface during the construction process, call the iterative training module to process the image data, match and compare it with the AI ​​acceptance sample model database, and output "qualified" or "unqualified" opinions through the data interface module to guide users to complete project self-inspection; the two interfaces collect WebService interface methods.

[0013] Moreover, the AI ​​acceptance database model database includes transformer body model data, installation and grounding model data, lightning arrester model data, sign warning sign model data, safety tool model data, electrical and mechanical interlocking model data, box station model data, box station shell model data, voltage transformer and current transformer model data.

[0014] The second object of the present invention is achieved by the following technical solution:

[0015] A processing method using the above-mentioned AI system for power grid expansion and installation service, the processing process is as follows:

[0016] When used on site, first use the business system module to upload photos of the electrical equipment and drawings to be accepted. After the files are digitized by the business system module 1, they interact with the iterative training module through the data interface module, and the iterative training module outputs the results to the business system module through the data interface module.

[0017] Moreover, the iterative process includes the sample data processing process and the acceptance data processing process in parallel:

[0018] The sample data processing process is as follows: based on part of the qualified sample image library and part of the defective sample database, image features are extracted, the extracted image features are trained through the machine learning algorithm model to obtain the input-output mapping relationship model, and another part of the qualified sample image library and another part of the defective sample database are used as input to verify the input-output mapping relationship model to obtain the accuracy evaluation of the model;

[0019] The acceptance data processing process is as follows: upload the on-site construction pictures that need to be accepted through the corresponding interface of the business system module. After the pictures are electronically processed, they are tested for image quality through the opencv js algorithm and the Yolov4 algorithm. The qualified pictures are segmented, and feature points are extracted from the segmented pictures. Then, the defects are automatically identified through the established input-output mapping relationship model to obtain the acceptance results, and the acceptance results are output to the business system module for storage and display.

[0020] The advantages and positive effects of the present invention are:

[0021] 1. The AI ​​system and processing method for power grid expansion and installation services of the present invention only requires customers to upload designated photos. After the pictures are electronically processed and processed by the iterative training module, the acceptance results can be output. There is no need for the customer manager to go to the site. It is easy to use and can greatly improve the efficiency of business expansion and installation acceptance.

[0022] 2. The AI ​​system of the present invention for the expansion and registration of power grid business can be operated independently or embedded in the existing State Grid Energy Internet Marketing Service (Marketing 2.0) system. Compared with the traditional manual mode of "acceptance-rectification-acceptance", the use of the present invention can reduce the processing time of each link by 67% and shorten the processing time of each link by 5 days on average. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 This is a diagram of the AI ​​system module structure for the power grid industry expansion and installation service of the present invention;

[0024] Figure 2 It is a module structure diagram of the business system of the present invention;

[0025] Figure 3 It is a diagram of the data interface module of the present invention;

[0026] Figure 4 is a reference graph of the contrastive learning strategy of iterative training of the present invention;

[0027] Figure 5 It is a multi-level accurate classification algorithm diagram of abnormal types of iterative training of the present invention;

[0028] Figure 6 It is a workflow diagram of the iterative training module of the present invention. DETAILED DESCRIPTION

[0029] The structure of the present invention is further described below with reference to the accompanying drawings and by way of examples. It should be noted that the present examples are descriptive rather than restrictive.

[0030] An AI system for power grid expansion and installation services, see Figure 1-Figure 6 , its invention point is: including AI acceptance sample model database, iterative training module, business system module and data interface module. The AI ​​acceptance sample model database includes multiple model sets such as qualified sample library and defective sample library, each model set stores multiple AI samples of the same type, the AI ​​acceptance samples are annotated by senior experts, and the AI ​​acceptance sample model database is used to provide standard data for iterative training for the iterative training module. The iterative training module provides iterative training for the system based on the model set in the AI ​​acceptance model database through algorithms such as OPEN AI and multi-scale feature extraction network, establishes a mapping relationship, and processes the submitted acceptance image data according to the established mapping relationship to determine whether the submitted acceptance data is qualified. The business system module can be developed based on the original system or run independently, and is used to provide digital support for acceptance data of multiple industries such as roads and bridges, water conservancy, fire protection, and planning; the data interface module standardizes the input acceptance data and outputs it to the smooth iteration module; the iterative training module calculates the results online, and outputs them to the business system module through the data interface module to realize the storage and feedback of whether the acceptance results are qualified.

[0031] The business system module 1, see Figure 2 , including the marketing business system and the online State Grid mobile power management system, the two can achieve data synchronization through the data middle platform.

[0032] The data interface module 2, see Figure 3 , including the "E-Assistant" module of the online State Grid mobile power management system, and an image data upload interface is added to this module. Users can upload the installation status of on-site electrical equipment through this interface during the construction process, call the iterative training module for image data processing, match and compare with the AI ​​acceptance sample model database, and output "qualified" or "unqualified" opinions through the data interface module to guide users to complete project self-inspection. The overall compliance with the State Grid SG-ERP architecture, the collection WebService interface mode.

[0033] The iterative model training 3, see Figure 4 and Figure 5 It is used to augment the AI ​​acceptance sample model database with abnormal samples through adversarial networks, target image fusion and other technologies, adopts feature extraction based on attention mechanism to adaptively realize efficient detection of target abnormal areas, and uses contrastive learning strategy to obtain better scene migration capabilities.

[0034] The AI ​​acceptance database model database 4 includes electrical equipment image samples of network-related equipment, which are divided into qualified sample libraries and defective sample libraries. For example: transformer body model (check whether the nameplate information is consistent with the applied capacity information), installation and grounding model (check whether the transformer shell is grounded, and whether the connection between the grounding lead and the main grounding electrode meets the specification requirements), lightning arrester model (check whether the lightning arrester grounding wire is independently led out for grounding), sign warning sign model (check whether signs such as "high voltage danger, no climbing" and "no throwing" are hung), safety tool model (whether the configuration is qualified and complete), electrical and mechanical interlocking model (check whether the two low-voltage main switches and bus tie switches of dual power customers are equipped with electrical and mechanical interlocking), box station model (whether the land around the box station is flat and whether there is a drainage channel), box station shell model (whether the shell is reliably grounded, and whether the resistance value of the main grounding resistance meets the technical specifications), voltage transformer and current transformer model (whether the body is intact, whether the leads, contacts and hardware are intact, and whether the connection is firm and free of rust), etc.

[0035] The processing method of the AI ​​system for power grid expansion and installation business is as follows:

[0036] When used on site, first use the business system module 1 to upload photos of the electrical equipment and drawings for acceptance. After the business system module 1 digitizes the files, it interacts with the iterative training module 3 through the data interface module 2, and the iterative training module 3 outputs the results to the business system module 1 through the data interface module 2. The iterative training module 3 can continuously optimize and iterate by calling the AI ​​acceptance model database 4.

[0037] The iterative process includes the sample data processing and the acceptance data processing in parallel:

[0038] The sample data processing process includes: extracting image features based on part of the qualified sample image library and part of the defective sample database, training the extracted image features through the machine learning algorithm model, obtaining the input-output mapping relationship model, using another part of the qualified sample image library and another part of the defective sample database as input, verifying the obtained input-output mapping relationship model, and obtaining the accuracy evaluation of the model;

[0039] The acceptance data processing process includes: uploading the on-site construction pictures that need to be accepted through the corresponding interface of the business system module, and after the pictures are electronically processed, image quality detection is performed through the opencv js algorithm and Yolov4 algorithm. The qualified pictures are segmented, and feature points are extracted from the segmented pictures. Then, defects are automatically identified through the established input-output mapping relationship model to obtain acceptance results, and the acceptance results are output to the business system module for storage and feedback.

[0040] Images that fail the image quality test will be fed back to the business system module and need to be re-uploaded. Unqualified images include: uploaded image data does not match the AI ​​data model sample (such as alternative equipment such as fruits and furniture), uploaded image data is blank or blurred, uploaded image resolution is not enough (such as long-range shooting), etc.

[0041] Although the embodiments and drawings of the present invention are disclosed for illustrative purposes, those skilled in the art will appreciate that various substitutions, changes and modifications are possible without departing from the spirit and scope of the present invention and the appended claims. Therefore, the scope of the present invention is not limited to the contents disclosed in the embodiments and drawings.

Claims

1. An AI system for power grid expansion and installation services, characterized by: Includes AI acceptance sample model database, iterative training module, business system module and data interface module; The AI ​​acceptance sample model database includes multiple model sets such as a qualified sample library and a defective sample library. Each model set stores multiple AI samples of the same type. The AI ​​acceptance samples are annotated with knowledge by senior experts. The AI ​​acceptance sample model database is used to provide standard data for iterative training for the iterative training module. The iterative training module provides iterative training for the system based on the model set in the AI ​​acceptance model database through OPEN AI, multi-scale feature extraction network, class activation mapping and other algorithms, establishes a mapping relationship, and processes the acceptance image data submitted by the user according to the established mapping relationship to determine whether the submitted acceptance data is qualified; The business system module is used to provide an input port for acceptance document data; The data interface module complies with the power grid network security strategy, and the data interface module performs standardization processing on the input acceptance image data, pre-processes the image content and image quality, and outputs it to the iterative training module; The iterative training module calculates the results online and outputs them to the business system module through the data interface module, so as to store and display whether the acceptance results are qualified.

2. The AI ​​system for power grid expansion and installation services according to claim 1 is characterized by: The business system module includes a marketing business system and an online State Grid mobile electricity management system, and the two can achieve data synchronization through a data middle platform.

3. The AI ​​system for power grid expansion and installation services according to claim 1 is characterized by: The data interface module includes the "E-Assistant for Electricity Management" module of the online State Grid mobile electricity management system, and an image data upload interface is added to the module. Users can upload the installation status of on-site electrical equipment through the interface during the construction process, call the iterative training module to process the image data, match and compare with the AI ​​acceptance sample model database, and output "qualified" or "unqualified" opinions through the data interface module to guide users to complete project self-inspection; the two interfaces collect WebService interface methods.

4. The AI ​​system for power grid expansion and installation services according to claim 1 is characterized by: The AI ​​acceptance database model database includes transformer body model data, installation and grounding model data, lightning arrester model data, sign warning sign model data, safety tool model data, electrical and mechanical interlocking model data, box station model data, box station shell model data, voltage transformer and current transformer model data.

5. A processing method of the AI ​​system for power grid expansion and installation service based on claims 1-4, characterized in that: The processing process is as follows: When used on site, first use the business system module to upload photos of the electrical equipment and drawings to be accepted. After the business system module digitizes the files, it interacts with the iterative training module through the data interface module, and the iterative training module outputs the results to the business system module through the data interface module.

6. The processing method of the AI ​​system for power grid expansion and installation services according to claim 5 based on claims 1-4 is characterized in that: The sample data processing process is as follows: based on part of the qualified sample image library and part of the defective sample database, image features are extracted, the extracted image features are trained through the machine learning algorithm model to obtain the input-output mapping relationship model, and another part of the qualified sample image library and another part of the defective sample database are used as input to verify the input-output mapping relationship model to obtain the accuracy evaluation of the model; The acceptance data processing process is as follows: upload the on-site construction pictures that need to be accepted through the corresponding interface of the business system module. After the pictures are electronically processed, they are tested for image quality through the opencv js algorithm and the Yolov4 algorithm. The qualified pictures are segmented, and feature points are extracted from the segmented pictures. Then, the defects are automatically identified through the established input-output mapping relationship model to obtain the acceptance results, and the acceptance results are output to the business system module for storage and display.

Images