Search

Insulators are critical components of transmission lines but are prone to failures that can jeopardize the safe operation of electrical power systems. Accurate detection of insulator defects is essential for timely maintenance. With advancements in object detection algorithm and artificial intelligence, insulator defect detection has garnered significant attention. However, detection accuracy remains an issue. To address this, we propose an improved transfer learning model. Our approach incorporates the Mish activation function and a global context network module to enhance the model's performance. The improved YOLOv9 model is trained and tested using two public datasets: the insulator defect image dataset and the China power line insulator dataset. Experimental results demonstrate that our model achieves optimal detection precision and recall rates of 99.84 and 99.92%, respectively—improvements of 1.06 and 1.09% over the actual YOLOv9. Additionally, our model outperforms other algorithms, such as RTDETR and SSD, particularly in adapting to complex backgrounds and detecting small targets. [ABSTRACT FROM AUTHOR]

To enhance the safety of hybrid inspection robots (HIRs) landing on power transmission lines (PTLs) with inclination and flexibility, this research derives a coupled dynamic model for a developed flying–walking power transmission line inspection robot (FPTLIR) to analyze the dynamic behavior of the FPTLIR during the landing process. The model uses the absolute nodal coordinate formulation (ANCF) for the dynamics of the PTL and the Hunt–Crossley theory for the contact model, integrating these components with the Euler–Lagrange method. A modular simulation was conducted to evaluate the effects of different landing positions and robot masses. An experimental platform was designed to evaluate the landing performance and validate the model, which confirms the method's accuracy, with a mean relative Z-displacement error of 0.004. Simulation results indicate that Z-displacement decreases with increased landing distance, with the farthest point showing only 34.4% of the Z-displacement observed at the closest point. Conversely, roll increases, with the closest point exhibiting 3.7% of the roll at the farthest point. Both Z-displacement and roll are directly correlated with the robot's mass; the lightest robot's Z-displacement and roll are 9.2% and 12.8% of those of the heaviest robot, highlighting the safety advantage of lighter robots. This research enables precise analysis and prediction of the system's responses during the landing process, providing significant insights for safe landing and design. [ABSTRACT FROM AUTHOR]

The modes of reactive power compensation in the power supply system powered by a synchronous generator through a cascade of intermediate transformers have been studied. It is shown that the reactive component of the common complex resistance of the generator and the power transmission path can cause relatively large levels of reactive power in the system. Experiments show that this power remains in the system even after the application of transverse capacitive compensation, which is not able to provide full compensation of reactive power in a power supply system with a cascade of transformers. Under these conditions, the optimal value of the capacitance of the transverse compensation capacitor was found using the search optimization method on the computer model. The results of the study show the possibility of achieving the mode of full compensation of reactive power due to a relatively small increase in the capacity of the compensating capacitor, which provides transverse compensation. [ABSTRACT FROM AUTHOR]

To enhance the safety of hybrid inspection robots (HIRs) landing on power transmission lines (PTLs) with inclination and flexibility, this research derives a coupled dynamic model for a developed flying–walking power transmission line inspection robot (FPTLIR) to analyze the dynamic behavior of the FPTLIR during the landing process. The model uses the absolute nodal coordinate formulation (ANCF) for the dynamics of the PTL and the Hunt–Crossley theory for the contact model, integrating these components with the Euler–Lagrange method. A modular simulation was conducted to evaluate the effects of different landing positions and robot masses. An experimental platform was designed to evaluate the landing performance and validate the model, which confirms the method’s accuracy, with a mean relative Z-displacement error of 0.004. Simulation results indicate that Z-displacement decreases with increased landing distance, with the farthest point showing only 34.4% of the Z-displacement observed at the closest point. Conversely, roll increases, with the closest point exhibiting 3.7% of the roll at the farthest point. Both Z-displacement and roll are directly correlated with the robot’s mass; the lightest robot’s Z-displacement and roll are 9.2% and 12.8% of those of the heaviest robot, highlighting the safety advantage of lighter robots. This research enables precise analysis and prediction of the system’s responses during the landing process, providing significant insights for safe landing and design.