基于计算机视觉的远程变电站状态监测与诊断新策略

REMOTE STATE MONITORINGAND DIAGNOSIS OF SUBSTATION BASED ON COMPUTER VISION

提出了一种基于计算机视觉的变电站畸变状态监测与诊断新策略,即通过与历史监测图像相比较对变电站的状态进行分类,确定是否传送图像和告警信号,调度中心接到告警信号后才需观察图像及处理异常情况。文章首先详细介绍了监测诊断系统的总体结构, 讨论了基于远程图像识别的电力设备状态诊断过程,然后结合图像的预处理过程,阐述了基于先验信息的电力仪表显示值的确定方法,并提出了适当的优化算法。通过对监测策略及图像识别的优化,建立的基于计算机视觉的变电站远程状态诊断系统降低了对通信信道的要求,并显著减轻了调度员连续观察大量实时图像的负担。

A new computer vision based strategy for substation distortion state monitoring and diagnosis is proposed, i.e., the current state of substation is classified by comparison between the monitoring images of current substation state and the historical monitoring images of substation to determine whether the images and alarm signals should be transmitted. When alarm signal is received, dispatching center observes the images and processes the abnormal situation. The architecture of the proposed monitoring and diagnosis system are described in detail, the power equipment state diagnosis process based on remote image recognition is discussed. Then combining with the preprocessing of images, the prior information based determination method for readouts of electric instruments is expounded and appropriate optimizing algorithm is put forward. Through the optimization of monitoring strategy and image recognition, the requirement to the communication channel for this computer vision based substation state diagnosis system is reduced, and the dispatchers' burden to continuously observe lots of real-time images is remarkably lightened.