摘要
为了提高风电场升压站巡检机器人的安全识别精度,实现设备巡检的智能化和无人化,提出了结合深度轻卷积和特征感知采样的改进目标检测算法,在此基础上结合巡检机器人设计了风电场升压站安全识别方法。实验结果显示,改进算法在闪络、老化、破损和短路这4类故障目标的检测识别中均表现出色,其检测准确率、召回率、F1分数与其他算法相比分别提高了6.31%、7.05%和8.34%,检测消耗的内存降低了23.73%。结果表明,所设计目标检测算法提高了巡检机器人对复杂、微小目标的检测精度和检测效率,为风电场的安全运行提供了有力保障。
In order to improve the safety recognition accuracy of the inspection robot of wind farm booster station and achieve the intelligence and unmanned inspection of equipment,the study proposes an improved target detection algorithm that combines deep light convolution and feature-aware sampling,on which a safety recognition method for the wind farm booster station is designed in combination with the inspection robot.The experimental results show that the improved algorithm performs well in the detection and identification of all four types of fault targets,namely flashover,aging,breakage and short circuit,and its detection accuracy,recall and F1 score are improved by 6.31%,7.05% and 8.34%,respectively,compared with other algorithms,and the memory consumed by detection is reduced by 23.73%.The results show that the target detection algorithm designed in the study improves the detection precision and detection efficiency of the inspection robot on complex and tiny targets,which provides a strong guarantee for the safe operation of wind farms.
作者
张洪彦
ZHANG Hongyan(Tuoketuo New Energy Business Department of Datang International Power Generation Co.,Ltd.,Hohhot,Inner Mongolia 010206,China)
出处
《自动化应用》
2025年第16期12-14,共3页
Automation Application
