Distribution network transformer real-time detection method and system based on neural computing rod

Abstract

The invention discloses a distribution network transformer real-time detection method and system based on a neural computing rod. Comprising the following steps: acquiring a to-be-detected distribution network transformer picture by using a yo3 target detection strategy, converting the to-be-detected distribution network transformer picture into image data, transmitting the image data to an image sensor for affine transformation, and outputting a corresponding actual position; acquiring transformer oil characteristic gas in a transformer by using a gas chromatographic analysis method, and carrying out pretreatment by fusing a WSMOTE strategy; constructing a positioning model based on a distributed sequential gas leakage source position evaluation strategy to detect the actual position of the distribution network transformer and the preprocessed characteristic gas, accelerating the positioning judgment process of the positioning model in real time in combination with an AI accelerator, and outputting a positioning result; and importing a positioning result into a Bayesian analysis model for secondary verification and judgment to obtain position information of the distribution network transformer, and displaying the result in real time by using a mobile terminal or a Raspberry Pi display. The neural computing rod of the AI accelerator is adopted to achieve the high-speed AI inference operation capability and the low-power-consumption capability.

Description

Technical field [0001] The present invention relates to the field of artificial intelligence, and more particularly to a real-time detection method and system for distribution transformers based on neural calculated rods. Background technique [0002] In order to achieve the intelligent management of the main power equipment of the substation, the safe and reliable operation of the power system is guaranteed, and the electric power company has strengthened the operation monitoring method and means of power equipment. At present, based on infrared thermal imagery, drone inspection, video online monitoring, etc., which produces massive visible light and infrared images, but mainly to handle various problems in the main equipment. This method will lead to a lot of human resources, and will result in an error due to lack of objectivity. Therefore, the use of image recognition techniques for power equipment target intelligence identification is necessary for subsequent power devices' ...

AI Technical Summary

Problems solved by technology

[0003] At present, artificial intelligence is gradually applied in the electric power industry. For the process of transformer identification in distribution network, the traditional solution is offline identification through image processing, but the identification effect is poor. There are also deep learning methods but the model is deployed in On the GPU server, the cost is relatively high, and the online service AI interface provided can be used when the network is good, but in certain usage scenarios, the online service AI interface cannot be used when there is no network or high network delay, so it is urgent to propose A portable embedded device with high performance, low power consumption and small size to complete the task of offline real-time identification of distribution network transformers

Images