Abstract: UHVDC valve elements are numerous and hard to monitor in real time. An on-line monitoring method is thus proposed for converter valve element conditions based on image recognition technology. The paper introduces the method to extract and pre-process image data, and analyzes the image characteristics of thyristor gate wire shedding and nut torque line displacement. The image recognition technology is adopted to realize intelligent detection of faults such as thyristor gate wire shedding and nut displacement. A prototype is designed for on-line monitoring of converter valve element conditions based on the image recognition and fault feature extraction method, and its software and hardware design methods are introduced in detail. The testing environment is set up on a valve tower of UHVDC to test the image analysis and fault feature extraction method. The testing results show that the proposed method can effectively identify the abnormal working conditions of the valve elements, and realize the early warning of their abnormal operations, and can prevent the further expansion of faults.

Key words: converter valve, image recognition technology, on-line monitoring, intelligent detection, early-warning