主管：中华人民共和国应急管理部
主办：应急管理部天津消防研究所
ISSN 1009-0029  CN 12-1311/TU
消防科学与技术

消防科学与技术 ›› 2021, Vol. 40 ›› Issue (3): 422-425.

• • 上一篇    下一篇

基于边缘计算的超特高压输变电走廊山火预警方法

赵钰，姜凌霄   

  1. 国网山西省电力公司检修公司，山西太原030032
  • 出版日期:2021-03-15 发布日期:2021-03-15
  • 作者简介:赵钰（1990-），男，国网山西省电力公司检修公司变电运维专业工程师，硕士，主要从事超特高压输变电设备运维工作，山西省太原市小店区高新开发区佳华街9 号，030032。
  • 基金资助:

Mountain fire early warning method in extraultra high voltage transmission and transformation corridor based on edge computing

ZHAO Yu，JIANG Ling-xiao   

  1. State Grid Maintenance Co.of SXPC, Shanxi Taiyuan 030032, China
  • Online:2021-03-15 Published:2021-03-15

摘要：

为克服现有输变电走廊山火预警技术的不足，提高对山火识别的抗干扰能力，提出一种基于边缘计算的山火预警方法。利用层次分析法，建立边缘监控节点位置选取的影响要素分析模型，通过判别矩阵确定每一要素的权重，根据最优解确定边缘监控节点的选取位置。针对输变电走廊山火识别环境，对边缘计算架构进行了详细设计，提高对输变电走廊的山火预警水平。为降低边缘卷积计算带来的算力开销，通过建立Mobilenet-SSD模型来训练边缘监控节点的山火图像识别数据，增加抗干扰措施，并验证模型效果。实验结果表明，训练模型适用于边缘计算架构且对输变电走廊山火能够有效识别。

关键词: 输变电走廊, 山火预警, 边缘计算, Mobilenet

Abstract: In order to overcome the shortcomings of the existing mountain fire early warning technology in transmission and transformation corridor and improve the anti- interference ability of mountain fire identification, a method of mountain fire early warning based on edge computing is proposed. Based on the analytic hierarchy process (AHP), an analysis model of influencing factors for the location selection of edge monitoring nodes is established. The weight of each element is determined by the discriminant matrix, and the location of edge monitoring node is determined according to the optimal solution. Aiming at the mountain fire identification environment in the transmission and transformation corridor, the edge computing architecture is designed in detail to improve the level of mountain fire early warning in the transmission and transformation corridor. In order to reduce the computational cost of edge convolution calculation, the Mobilenet-SSD model is established to train the fire image recognition data of edge monitoring nodes, and anti-interference measures are added to verify the model effect. The experimental results show that the training model is suitable for the edge computing architecture and can effectively identify the fire in the transmission and transformation corridor. 

Key words:  , power transmission and transformation corridor, mountain fire early warning, edge computing, Mobilenet