基于改进模式识别的无人值守风电场群组机器人集中巡检研究  

Research on centralized inspection of unmanned wind farm group robots based on improved pattern recognition

作　　者：董礼 程丽敏 赵博 王雁冰 商志强 朱盼盼 Dong Li;Cheng Limin;Zhao Bo;Wang Yanbing;Shang Zhiqiang;Zhu Panpan(CGN Wind Power Co.,Ltd.,Beijing 100000,China;Beijing Jinfeng Huineng Technology Co.,Ltd.,Beijing 102600,China)

机构地区：[1]中广核风电有限公司,北京100000 [2]北京金风慧能技术有限公司,北京102600

出　　处：《可再生能源》2025年第3期346-352,共7页Renewable Energy Resources

基　　金：国家自然科学基金(51675498)。

摘　　要：由于风电场设备种类繁多、运行环境复杂多变,通常无人值守,故障难以及时发现。传统巡检方法耗时长且识别准确性低,导致故障处理不及时,影响风电场稳定运行和发电效率。为此,文章针对无人值守风电场群组提出了基于改进模式识别的机器人集中巡检方案。对于风电场群组变压器故障、设备温度异常和齿轮箱声音异常情况,分别利用BP神经网络算法、模糊模式识别算法和经验模态分解算法对其展开巡检,并在某大型风力发电场中对所提方法进行测试。结果表明,所提方法可实现对风电场群组中各类故障的巡检,第一时间获取到故障信号,避免了安全事故的发生;识别准确率在92.3%以上,召回率与F1分数也优于对比方法,表明本文方法在识别故障样本方面更为全面,能够有效地进行故障检测。Due to the wide variety of wind farm equipment and complex operating environment, it is usually unattended and difficult to find faults in time. The traditional inspection method takes a long time and has low identification accuracy. As a result, the fault is not handled in time, which affects the stable operation and power generation efficiency of wind farms. Therefore, a robot centralized inspection scheme based on improved pattern recognition is proposed for unattended wind farm groups. For transformer faults, equipment temperature anomalies and gearbox sound anomalies in wind farms, BP neural network algorithm, fuzzy pattern recognition algorithm and empirical mode decomposition algorithm are used to carry out inspection, and the proposed method is tested experimentally in a large wind power station. The results show that the proposed method can realize the inspection of various faults in wind farms. The first time to obtain the fault signal, to avoid the occurrence of security accidents;The recognition accuracy rate remains above92.3%, and the recall rate and F1 score are also better than the comparison method, indicating that the proposed method is more comprehensive in identifying fault samples and can detect faults more effectively.

关 键 词：改进模式识别 BP神经网络算法 经验模态分解算法 齿轮箱声音异常 变压器故障 

分 类 号：TK81[动力工程及工程热物理—流体机械及工程]