Rong Jia

Heavy construction machinery is essential for power infrastructure construction and maintenance, making safety distance measurement crucial for ensuring operational safety. Existing methods, which rely on target space mapping, face challenges such as error accumulation and limited accuracy. Motivated by this, this paper proposes a refined measurement method using high‐precision, anti‐interference LiDAR as the 3D data source. The method employs multi‐layer set abstraction for 3D feature extraction and introduces instance‐aware downsampling to homogenise point cloud features across targets of varying distances and sizes. Subsequently, an instance centroid prediction module aggregates contextual information to estimate precise spatial positions. Ultimately, by integrating optimal representative point selection with a distance calculation mechanism, the method achieves accurate minimum safety distance measurement for near‐electricity operations. Experimental results demonstrate superior accuracy in target location identification and distance measurement in complex scenes, outperforming existing methods. This approach provides theoretical and practical support for risk prevention in near‐electricity operations of heavy construction machinery.

The environment of the power operation site is complex and changeable, and the accurate identification of the wearing status of workers’ safety helmets is significant to ensure personal safety and the stable operation of the power system. Existing research suffers from high rates of missed detections and limited ability to discriminate fine-grained states, especially the identification of “wrongly wearing” states. Therefore, this paper proposes an intelligent identification method of safety helmet status for power workers based on a dual-path attention network. We embed the convolutional block attention module (CBAM) in the two paths of the backbone and neck layers of YOLOv5 and enhance the feature focusing ability of the key areas of the helmet through the channel-spatial attention coordination, so as to suppress the interference of complex background. In addition, a special dataset covering power scenarios is constructed, including fine-grained state annotation under various lighting, different poses, and occlusion conditions to improve the generalization of the model. Finally, the proposed method is applied to the images of the electric power operation site for experimental verification. The experimental results show that the proposed YOLO-CBAM achieves an outstanding mean average precision of 98.81% for identifying all helmet states, providing reliable technical support for intelligent safety monitoring.

Construction machinery has been widely used to upgrade the power industry. Nevertheless, high-voltage and high-current operational conditions necessitate stringent safety distance protocols. Simultaneously, the operational machinery autonomy and concentration of live equipment impose considerable constraints on current distance measurement techniques. To address these challenges, we propose a refined measurement method for the safety distance of construction machinery based on binocular vision and target state perception. First, the Edge-Segformer model is employed to achieve fine-grained detection of construction machinery and other targets, dynamic recognition of operational states, and precise segmentation of operational zones. Next, the IGEV-Stereo model is utilized to obtain high-precision disparity maps, and the mapping relationship from pixel coordinates to 3D space is established based on the triangulation principle. Finally, through local contour analysis and depth information, the key measurement points for the minimum safety distance (MSD) are located, and the shortest Euclidean distance from a key point to the boundary of the operational zone is calculated, serving as the MSD during machinery operation. The experimental results demonstrate that our method achieves a fine-grained detection accuracy of 92.80 %, exhibiting a 3D reconstruction error of 3.23 % and MSD measurement error of 6.8 %, validating its effectiveness in various substation scenarios.

Power production is a complex process that involves multiple interactions, which require rich semantic knowledge to categorize and evaluate. Utilizing high-level image understanding to accurately identify risks is significant in ensuring staff safety and grid stability. In this work, we propose image captioning and semantic text similarity analysis (IC-STSA), a risk identification method for complex power operation scenarios the employs image captioning and semantic text similarity analysis. First, image caption is used to transform an image at the site into the text that best describes the visual data. Next, a safe power production (SPP) knowledge base is established, after which semantic text similarity analysis is introduced to transform the risk identification into a text classification problem. By comparing the image caption information and the SPP knowledge base and identifying similarities, we achieve refined descriptions of operation scenarios and thereby risk identification. In the final step, actual images are used to verify the findings of the IC-STSA. Results show the proposed method performs well, generating accurate, targeted descriptions, and risk identification.

To enhance the accommodation capacity of renewable energy and promote the coordinated development of multiple energy, this paper proposes a novel economic dispatch method for an integrated electricity–heat–hydrogen energy system on the basis of coupling three energy flows. Firstly, we develop a mathematical model for the hydrogen energy system, including hydrogen production, storage, and hydrogen fuel cells. Additionally, a multi-device combined heat and power system is constructed, incorporating gas boilers, waste heat boilers, gas turbines, methanation reactors, thermal storage tanks, batteries, and gas storage tanks. Secondly, to further strengthen the carbon reduction advantages, the economic dispatch model incorporates the power-to-gas process and carbon trading mechanisms, giving rise to minimizing energy purchase costs, energy curtailment penalties, carbon trading costs, equipment operation, and maintenance costs. The model is linearized to ensure a global optimal solution. Finally, the experimental results validate the effectiveness and superiority of the proposed model. The integration of electricity–hydrogen coupling devices improves the utilization rate of renewable energy generation and reduces the total system operating costs and carbon trading costs. The use of a tiered carbon trading mechanism decreases natural gas consumption and carbon emissions, contributing to energy conservation and emission reduction.

In the context of global industrialization, integrated energy systems (IES) have emerged as a crucial means of facilitating green and low-carbon transformation. To comprehensively explore the potential for carbon reduction through source-load synergy and achieve low-carbon and high-efficiency operation of the system, an IES optimal scheduling method that incorporates a reward and punishment mechanism for dynamic electricity and thermal carbon emission factors is proposed. Firstly, the concepts of energy coupling and thermal energy cascade utilization are considered, and an IES-coupled equipment cascade utilization model with carbon capture is established with the objective of realizing high-grade use of thermal energy in combination with thermal load demand. Secondly, an improved IES dynamic electricity and thermal carbon emission factors model is proposed to characterize the carbon emission intensity and renewable energy consumption capacity. Based on this, an integrated demand response (IDR) mechanism driven by electricity, thermal reward, and carbon price is established to guide users toward low-carbon behaviors and renewable energy consumption. Finally, a two-stage optimal scheduling model is constructed. In the pre-scheduling phase, the initial scheduling plan is solved, and the load distribution is adjusted by the integrated demand response. In the re-scheduling phase, the load curve is updated to obtain the final scheduling plan. Through simulation verification, the proposed scheduling model reduces carbon emissions by 7.7 %, improves the renewable energy consumption rate by 4 %, and reduces the comprehensive operating cost of the system by 4.3 %, which indicates that the method proposed in this paper can effectively promote the emission reduction and consumption and reduce the operating cost of the system.

With the increasing complexity of the distribution network structure, enhancing the efficiency and reliability during fault restoration has become a focal point. Based on the multi-source information collected by traditional sensors, such as CT and PT, and intelligent sensors, such as D-PMU, and the graph calculation model, the fault recovery problem of a multi-objective distribution network is studied. Firstly, a power flow calculation model and operation constraint adaptable to topology changes are proposed under the graph calculation framework. The minimum spanning tree theory is utilized to define the blackout range and recovery path set. Secondly, the intelligent sensor D-PMU is configured to collect fault information to ensure that at least one of any two connected load vertices is configured with D-PMU. Thirdly, a topological evolution model is established that considers repeated primary and secondary transfer in outage areas while exploring possible recovery strategies deeply. Finally, a distribution network in Shaanxi Province is taken as an example to verify the model. The experiment shows that the strategy in this paper dynamically adjusts the recovery strategy through four means—one transfer, one repeat transfer in the outage area, two transfers, and cutting off part of the outage load—and the overall recovery rate is increased by more than 20%.

Under the carbon peaking and carbon neutrality target background, efficient collaborative scheduling between distribution networks and multi-microgrids is of great significance for enhancing renewable energy accommodation and ensuring stable system operation. Therefore, this paper proposes a collaborative optimization method for the operation of distribution networks and multi-microgrids with shared energy storage based on a multi-body game. The method is modeled and solved in two stages. In the first stage, a multi-objective optimization configuration model for shared energy storage among multi-microgrids is established, with optimization objectives balancing the randomness of renewable energy fluctuations and the economics of each microgrid undertaking shared energy storage. The charging and discharging interactive power of energy storage and each microgrid at various time periods are obtained and passed to the second stage. In the second stage, with the distribution network as the leader and shared energy storage and multi-microgrids as followers, a game optimization model with one leader and 2 followers is established. The model is solved based on an outer-layer genetic algorithm nested with an inner-layer solver to determine the electricity purchase and sale prices among the distribution network, multi-microgrids, and shared energy storage at various time periods, thereby minimizing operational costs. Finally, based on the power interaction of microgrids to measure their contributions, an improved Shapley value cost allocation method is proposed, effectively achieving a balanced distribution of benefits among the distribution network, shared energy storage, and multi-microgrids, thereby improving overall operational revenue. Meanwhile, a new method for calculating the shared energy storage capacity and the upper limit of charging and discharging power based on a game framework was proposed, which can save 37.23% of the power upper limit and 44.89% of the capacity upper limit, effectively saving the power upper limit and capacity upper limit.