Research On Substation Equipment Safety Early Warning Technology Based On Infrared Imaging

With the development of power system in the smart grid, the ability for substations andother electrical equipment on-line monitoring and safety warning to detect faults is required.However, the current substation fault detection technology cannot achieve alarmsautomatically in a timely manner for equipment failure. With the combination of substationonline monitoring system and image analysis based on infrared imaging and video analysistechniques, it can be achieved real-time alarm event detection, which can greatly enhance thesafe operation of the substation emergency support capabilities, which has importanttheoretical significance and engineering value.With infrared cameras remote temperature monitoring system, automatic detection andtimely fault alarm cannot be realized. A substation electrical equipment infrared remoteviewing and safety warning system is thus proposed, which can automatically capture infraredtemperature cycling image substation electrical equipment with infrared and visible camerasmounted on the head. For the problems of remote viewing images in complex background andlarger changes in the environment, a method of object segmentation is put forward, which isinitialized object segmentation for images interested in the device. It can effectively solve themost difficult problems extracting equipment and the later image analysis can be proceedsmoothly, which is the benefit for the database realization, fault diagnosis and expert systemalarming.Aiming at problems of infrared image acquisition system in Position offset, rotationangle and geometric distortion, a registration method based on phase correlation and Harriscorner matching combining images is adopted to achieve a deviation for most imageregistration, and yet with the phenomenon of individual devices beneath figure. With deeplyanalysis of mismatch causes, an improved SIFT feature registration method is then put forward according to phase correlation method of direction of the main movement to reducethe search area, which can reduce the matching time as well as promoting matching accuracy.The experiments demonstrate that the method has higher registration accuracy to achievesub-pixel level and stronger adaptability to successfully registration of all test images, and thereal-time requirements can be satisfied.The oil and transformer and oil pillow radiator are important need for real-timemonitoring of substation equipment. The lower or higher of oil level or barrier on a heat sinkthe oil can cause serious security risks. Aiming at the problem of lower infrared imageresolution of oil pillow, an improved watershed segmentation algorithm using different scalesof structural elements to calculate the gradient to enlarge the weak edge image in varyingdegrees is proposed. At the same time, it adopts minimal value forced calibration technologyto effectively prevent over-segmentation. The experiments show that the algorithm canaccurately and effectively split the oil industry and the oil pillow and accuracy can meet thesystem requirements. As the transformer radiator fault boundary is not clear, an improvedrobust fuzzy clustering algorithm is proposed. Nuclear methods and spatial and graysimilarity measure are introduced to enhance the robustness of the algorithm. The experimentsshow that the split effect is better than fuzzy clustering segmentation FCM, KFCM, etc.,which can clearly divide faulty heat sink and location.The electrical equipment using SF6gas insulated has been put into a large number of gridoperations, and inflatable equipment in case of leakage occurrence will cause serious adverseconsequences. In order to achieve continuous monitoring and timely unmanned, an automaticleak detection method is proposed. With the dynamic characteristics similar SF6leakage ofsmoke in the infrared video, the methods of Gaussian mixture background modeling isadopted to SF6gas leak detection. Based on mathematical morphology Learn methods, thespots and other noise are removed to find the suspect. With analyzing the smoke suspect areaand extract fumes, it can be differentiated from the unique dynamic characteristics and thespill area is ultimately determined. Through the multiple imaging infrared gas leak collectedvideo SF6leak detection, the experimental results show that this method can effectivelyovercome background disturbance, such as the light changes, accurately realize the leak ofautomatic detection and positioning, which are feasible and effective for achieving leakautomatic monitoring.