To address the challenges of high missed detection rates for minute transmission line defects, strong complex background interference, and limited computational power on edge devices in UAV-assisted power line inspection, this paper proposes a lightweight improved YOLOv12 real-time detection model. First, a Bidirectional Weighted Feature Fusion Network (BiFPN) is introduced to enhance bidirectional interaction between shallow localization information and deep semantic features through learnable feature layer weighting, thereby improving detection sensitivity for line defects. Second, a Cross-stage Channel-Position Collaborative Attention (CPCA) module is embedded in the BiFPN’s cross-stage connections, jointly modeling channel feature significance and spatial contextual relationships to effectively suppress complex background noise from vegetation occlusion and metal reflections while enhancing defect feature representation. Furthermore, the backbone network is reconstructed using ShuffleNetV2’s channel rearrangement and grouped convolution strategies to reduce model complexity. Experimental results demonstrate that the improved model achieved 98.7% mAP@0.5 on our custom transmission line defect dataset, representing a 3.0% improvement over the baseline YOLOv12, with parameters compressed to 2.31M (8.3% reduction) and real-time detection speed reaching 142.7 FPS. This method effectively balances detection accuracy and inference efficiency, providing reliable technical support for unmanned intelligent inspection of transmission lines.

To mitigate black start failures resulting from energy storage state of charge (SOC) exceeding operational limits, this study develops a restoration strategy incorporating SOC constraints. Firstly, an adaptive SOC control without bias for energy storage units is proposed to achieve SOC balance. Secondly, the maximum power point tracking (MPPT) mode for photovoltaic power generation is integrated with the load tracking mode, enabling effective tracking of load variations when the photovoltaic output is sufficient; conversely, when the photovoltaic output is inadequate, the energy storage output compensates for the shortfall, thus avoiding the SOC limit due to an insufficient remaining SOC of the energy storage, while also significantly reducing the quantity of charge and discharge cycles undergone by the energy storage units. Finally, the simulation results show that, according to the proposed control strategy for recovery, the maximum system frequency of the black start process does not exceed 50.11, and the minimum is not lower than 49.82, which are within reasonable limits. At the same time, the number of charge/discharge conversions of the three storage batteries is 12 when the PV system adopts the coordinated control of MPPT and load tracking, and the number of charge/discharge conversions of the storage batteries in the PV MPPT mode is 21. This ensures the success of the black start process and prolongs the life of the energy storage battery.

In ultra-high-speed networks, identifying hosts with high source or destination cardinality, or large cardinality changes, provides compact measurements for network monitoring and security analysis. Currently, schemes for super-host identification have demonstrated good performance, and most of them target one type of super host. However, real networks contain multiple types of super hosts with different distributions. Supporting source cardinalities, destination cardinalities, and inter-cycle changes within one memory-bounded structure requires coordinating measurement and address recovery. To address this challenge, we present H-TD Sketch, a novel sketch structure that enables rapid and accurate identification of multiple super-host types under limited memory constraints by maintaining counts of both source and destination cardinalities. The H-shaped Bloom filter and information groups enable direct traceability of super-host addresses at the end of the measurement cycle. We conduct a theoretical analysis of H-TD Sketch, covering its space and time complexity and the accuracy of its cardinality estimates. We evaluate cardinality estimation and super-host identification on a CPU platform using real-world CAIDA traces. We also integrate H-TD Sketch into Open vSwitch and evaluate its packet-processing throughput using generated traffic. Our experimental findings demonstrate that H-TD Sketch exhibits superior accuracy and performance in both host-cardinality estimation and super-host identification tasks.

A sketch is a probabilistic data structure that can accurately estimate massive network traffic with a small memory overhead. To improve the measurement accuracy, most sketch-based schemes separate elephant flows from mouse flows to accommodate the skewed network traffic. However, the increased algorithmic complexity often results in a sacrifice of measurement throughput. In addition, some improved sketches may be over-reliant on the skewed distribution of traffic, which results in unstable accuracy. To this end, a novel sketch, called Air Sketch, is proposed in this paper. It treats flows of different sizes as air with different temperatures. Meanwhile, a deterministic replacement strategy is applied to elephant flows. In order to improve throughput, an asymmetric insertion and query algorithm with a global hash is designed. The performance of Air Sketch is evaluated using real traffic transaction datasets, anonymized internet traces, and synthetic Zipf datasets. The experimental results demonstrate that Air Sketch can outperform the best typical measurement methods by up to 27 times in flow size measurement and up to 40 times in elephant flow detection. Additionally, Air Sketch achieves high accuracy and stability while achieving high insertion and query throughput.