Fault diagnosis of power transmission systems (PTSs) is crucial for ensuring the reliability of power grids because most grids are exposed to harsh environments. For integrity diagnosis of PTSs, this paper proposes a nondestructive patrol inspection method that employs a smart inspection system (SIS) mounted on an unmanned aerial vehicle (UAV). This system overcomes the geographical limitations faced in accessing PTSs. The SIS includes an ultraviolet camera system that can detect partial discharges on the damaged surfaces of PTSs. The SIS is characterized by three main features. First, it employs an automatic line-tracking method based on image-processing methods. Second, defect locations are automatically estimated on the basis of the consistency of partial discharges, where consistency indicates the steadiness of corona discharges in one-dimensional global coordinates. Third, an alarm is triggered on the basis of the combination of the result from a statistical outlier detection method and the estimated partial discharge locations on a two-dimensional partial discharge map. Finally, a UAV equipped with the proposed system is field-tested for PTS inspection in the autopilot mode. Field tests conducted on ultrahigh-voltage PTSs confirmed the damage detection capabilities of the proposed method and demonstrated its effectiveness and convenience.

Transmission lines (TLs) are prone to frequent failures owing to their exposure to extreme environments. It is essential to monitor these failures and repair the TLs on time; moreover, their health must be accurately estimated to ensure reliability and safety. Therefore, this study proposes an intelligent monitoring method with novel sensors deployed on a UAV. Specifically, this paper presents a method to not only cognize TLs and their environments in real time but also estimate the sag by combining the measured point-cloud data from a Lidar with the flight information of a UAV. Environmental cognition in real time addresses coordinate transformation and probabilistic downsampling for voxelized mapping, thus ensuring limited hardware requirement. A robust random-sample consensus is introduced to effectively extract point-cloud data for TLs, thereby accurately estimating the sag. Field tests confirmed the accuracy of the proposed method and demonstrated the effectiveness and advantages of the smart environmental cognition system. One flight was sufficient for estimating the sag on both sides of the TLs, demonstrating the economic feasibility of the proposed method. Considering the inherent advantages of the diagnostics and prognostics of a UAV, the proposed method has potential applications in the reliable operation and proactive maintenance of TLs.