Study On Automatic Status-recognition Method Of Transformer-fan-set Based On Video Surveillance

In order to effectively prevent power-transformer-accident, an automatic status-recognition method of fan based on video surveillance is proposed to provide guarantee for the security of transformers. And a parameter-detecting method at different levels based on improved Hough transform for circle center parameters and radius parameters is put forward to detect power transformer fans in complicated backgrounds. Subsequently, Otsu image segmentation algorithm on the basis of genetic algorithm is applied in annular domains around accurate circle center data, and calculating the distribution proportion of fan-type pixels in annular domains in binary image as a state identification feature vectors. On the foundation of above, study on Machine Learning and its applications in status-recognition of fan are presented, and classifier fan state recognition based on Support Vector Machine (SVM) is designed and implemented. By collecting samples for SVM training, the optimal separating surface with the minimization structural risk is developed for real-time monitoring of the functioning of the state transformer fan.