查询结果:   虢韬,沈平,王伟,时磊,杨渊.基于大气电场值的雷电发生识别方法[J].计算机应用与软件,2018,35(2):86 - 90.
中文标题
基于大气电场值的雷电发生识别方法
发表栏目
应用技术与研究
摘要点击数
680
英文标题
LIGHTNING OCCURRENCE RECOGNITION METHOD BASED ON ATMOSPHERIC ELECTRIC FIELD VALUE
作 者
虢韬 沈平 王伟 时磊 杨渊 Guo Tao Shen Ping Wang Wei Shi Lei Yang Yuan
作者单位
贵州电网有限责任公司输电运行检修分公司 贵州 贵阳 550005 国网电力科学研究院武汉南瑞有限责任公司 湖北 武汉 430074 武汉大学计算机学院 湖北 武汉 430074   
英文单位
Transmission Line Operation and Maintenance Branch of Guizhou Power Grid Co.,Ltd.,Guiyang 550005,Guizhou,China Wuhan NARI Limited Liability Company,State Grid Electric Power Research Institute,Wuhan 430074,Hubei,China School of Computer,Wuhan University,Wuhan 430074,Hubei,China   
关键词
雷电 大气电场 小波包 随机森林
Keywords
Lightning Atmospheric electric field Wavelet packet Random forest
基金项目
湖北省自然科学基金项目(2014CFB194)
作者资料
虢韬,高工,主研领域:输电线路运维管理。沈平,工程师。王伟,工程师。时磊,助理工程师。杨渊,工程师。张磊,工程师。陈玥,工程师。胡有,学士。罗飞,讲师。 。
文章摘要
云地间雷电会对地面设施造成极大的破坏。对雷电的预警除了使用雷达,卫星等设备外,大气电场仪是一种有效、相对低廉的定点雷电监测预警设备。由于雷电不是唯一可引起大气电场值改变的因素,并且雷电的发生是瞬时的、离散的,现有基于大气电场值的雷电发生识别方法常出现假阳性情况。为了提高识别准确率,基于大气电场值,提出一种利用小波包和随机森林的雷电发生识别方法。区分特征上,该方法使用小波包提取雷电过程中大气电场值变化的频域值,并结合能量大小构建区分特征;分类器上,利用随机森林分类器精度高,可解释性强的特点对负样本和三个等级强度正样本进行区分。实验结果表明,该方法无论是在中、强等级正样本对负样本的雷电发生区分还是弱正样本对负样本的区分中均取得了非常高的正确率,证明了该方法的有效性。
Abstract
Lightning between cloud and ground can cause great damage to the ground facilities. For lightning monitoring and warning, besides of the radar and the satellite, the atmospheric electric field instrument is an effective, relatively inexpensive fixed-point instrument. Because lightning is not the only factor that can cause the change of atmospheric electric field value, and the occurrence of lightning is random and discrete, the existing lightning identification methods based on atmospheric electric field value often appears false positive situations. In order to improve the recognition accuracy, still based on the atmospheric electric field value, a method was proposed for distinguishing the true and false positive lightning by using wavelet packet and random forest. In aspect of features, wavelet packet was used to extract the frequency-domain value of atmospheric electric field during the thunder and lightning process, and the distinguishing feature was constructed based on the energy. In aspect of classifier, using the characteristics of random forest classifier with high precision and strong explanatory power, negative samples and three positive samples of intensity were distinguished. The experimental results showed that the method not only had a very high accuracy rate in the distinction between the negative samples and the samples in average and strong, but also the same performance was obtained in the distinction between the negative and weak samples, which proved the effectiveness of this method.
下载PDF全文