【摘要】 隔离开关长期使用容易合闸不到位,导致电力系统事故。合闸是否到位区别不明显，检测难度大，且一个供电局要检测数千个不同类型的隔离开关，这些难点使得目前的检测方法尚无规模化实际应用。提出一种基于云边融合架构的隔离开关精确检测系统，可大规模部署，成本较低实时性好；针对算法难以兼顾多类型开关的难题，将图像识别与硬件架构相结合，提出了算法选择器，显著提高了识别精度；给出3种方法对识别结果精确计算合闸角度，尤其是通过详细的理论推导，提出垂直拟合法，克服了经典线性拟合法的缺陷。该系统已得到实际应用。更多还原

【Abstract】 The long term use of isolation switches can easily lead to inadequate closing,resulting in power system accidents. There is an insignificant difference in whether the closing is in place,with high difficulty in detection; and a power supply bureau needs to monitor thousands of different types of the isolation switches. These difficulties make it difficult for large-scale practical applications at home and abroad. The paper propose a cloud-edge fusion hardware architecture with large-scale deployment potential for the precise detection of the isolation switches,realizing low cost and good real-time performance of the architecture. In response to the difficulty of taking into account multiple types of switches in algorithms,an algorithm selector is constructed by combining image recognition with hardware architecture,significantly improving the recognition accuracy. Three methods are used to accurately calculate the closing angle based on the recognition results,and the vertical fitting method is proposed to overcome the shortcomings of the classical linear fitting method especially through detailed theoretical derivation. The system has been practically applied.更多还原