Transformer substation smoke and fire intelligent identification monitoring method based on deep learning

Abstract

The invention discloses a transformer substation smoke and fire intelligent identification monitoring method based on deep learning. Improvement and optimization of a video recognition model and an image recognition model are carried out on an actual scene of a transformer substation; two improved model frameworks are fused; the disadvantages of the two improved model frameworks are avoided as much as possible, and respective advantages of the two improved model frameworks are developed, a reasonable and flexible detection method is designed, an image recognition model is used for monitoring at ordinary times, after smoke is detected, a video recognition model is automatically called for secondary rechecking, an alarm signal is sent to a monitoring platform after checking is accurate, anddetection and early warning work can be effectively completed.

Description

technical field [0001] The invention relates to a method for intelligent identification and monitoring of pyrotechnics in substations based on deep learning, and belongs to the technical field of artificial intelligence monitoring and identification for substation safety. Background technique [0002] Fire has always been one of the great threats to the safety of people's lives and properties. The suddenness, frequency and high destructive power of fires seriously threaten people's lives, properties and the natural environment. The substation is the hub of the power system. Once a fire breaks out, it may cause the collapse of the entire power grid system, seriously endangering the reliability of power supply. Therefore, doing a good job of fire prevention measures in substations is a crucial task to ensure the safe and stable operation of the power grid. The main reasons for fires in substations are as follows: electrical equipment causes fires, and substations contain a la...

AI Technical Summary

Problems solved by technology

The traditional sensor-based smoke detection method is easily affected by the environment, has a certain alarm delay and is not suitable for a wide range of environments such as outdoors

With the rapid development of intelligent video technology, research on smoke detection using image recognition or video recognition technology has begun. Most current applications use images alone for detection or video alone for detection, but these two methods have their limitations. Responsibility: Image recognition takes up less resources but is more likely to generate false positives; video recognition has higher accuracy but takes up too many resources. When network bandwidth resources are tight and multi-channel video real-time transmission is detected, the server load is heavy

Images