Development Of Fault Recognition Software Based On Multi-source Image For Electrical Equipment And Environmental Monitoring

Towards the demand of power system state monitoring and the problem of multi-type and multi-target co-occurrence of electrical equipment faults,multi-source image fault detection technology is deeply researched,which based on visible,infrared,and ultraviolet images.Typical fault detection algorithms such as overheating,discharge and open flame detection are studied.A remote fault detection system for electrical equipment with high scalability and portability is designed and developed,which provides a complete scheme for electrical equipment state monitoring.Aiming at the detection of three typical faults of electrical equipment overheating,flame and discharge under infrared and ultraviolet images,an overheating fault detection algorithm based on static temperature feature analysis are developed,flame and discharge fault detection algorithms based on dynamic feature analysis are developed.With reference to the infrared detection standard for electrical equipment faults,the algorithm flow of overheating fault detection is proposed.For flame and discharge detection,the corresponding dynamic feature type is selected,and the dynamic feature analysis method is designed.Furthermore,in view of the problem that traditional dynamic feature analysis algorithms are only suitable for single target,a multi-target fault detection method based on heuristic classification algorithm is proposed.The above algorithms solve the problem of multi-type and multi-target electrical equipment fault detection based on infrared and ultraviolet images.Since the custom RGB flame dataset has a small number of samples and only one detection target type,in order to improve the accuracy and generalization performance of the flame detection network,a flame detection algorithm with attention mechanism is proposed.The channel attention mechanism is introduced into the deep layer of the Yolov3 detection network,and the deep semantic dependence is captured by the weighted summation of the feature maps,so that the deep semantic information is focused on the flame target type,and the accuracy of flame detection is improved.This paper also developed a smoke recognition algorithm as a supplement to flame detection.Based on the above fault detection algorithms,an electrical equipment fault detection system software and hardware platform was designed and developed according to the existing monitoring hardware configuration and monitoring requirements of power production scenarios,including hardware platform,communication framework and software platform.The software platform specifically includes background fault detection system,monitoring client software and custom communication protocol module.Through the design of the software and hardware middle layers,the portability,flexibility and scalability of the system are improved.Finally,the custom datasets and simulated fault experiments were used to verify the detection algorithm and software and hardware platform,which realized the intelligent and automated monitoring of the state of electrical equipment in the power production scenario.