Assistant analysis system and method for intelligent robot patrol of substations

Abstract

The invention discloses an assistant analysis system and method for intelligent robot patrol of substations. The assistant analysis system comprises an image shooting servo module, a deep-learning identification module and abnormal state inference module; the image shooting servo module is used for assisting an intelligent robot in shooting sample images and inputting the sample images to the deep-learning identification module; the deep-learning identification module is used for establishing three-level pattern identification systems for identifying equipment, parts and statuses respectively,and each pattern identification system is used for object identification of the sample images and inputting identification results to the image shooting servo module and the abnormal state inferencemodule; the abnormal state inference module is used for performing inference engine on equipment status information identified by the deep-learning identification module and displaying inference results to users. Compared with a traditional characteristic-based image identification method, the assistant analysis method has the advantages of high accuracy and generalization capability.

Description

technical field [0001] The invention relates to the technical field of substation intelligent patrol inspection, and more specifically relates to a substation intelligent robot patrol auxiliary analysis system and method. Background technique [0002] Daily inspection is the most important means to ensure the safe and stable operation of substations. With the rapid development of society and economy, the investment in power grid construction has been increasing, and the contradiction between the supplement of technical personnel and the speed of power grid construction has become increasingly apparent. [0003] The traditional substation inspection work is carried out by substation site operators who go to the substation under their jurisdiction to do regular periodic inspections every day according to the formulated time node plan, and record the operation data of relevant equipment. In the context of the rapid development of electric power, there are many substations under...

AI Technical Summary

Problems solved by technology

The identification of traditional machine learning faults requires prior knowledge to design feature parameters. The selection of feature parameters will directly affect the effect of fault identification, and the captured images are affected by environmental factors such as shooting position, angle, and illumination, resulting in accurate fault identification. The volatility is very large, which cannot meet the requirements of practical applications

Images