基于随机森林算法的电线覆冰检测技术  

Wire Icing Detection Technology based on Random Forest Algorithm

作　　者：罗扬燚 卢会国[1] 蒋娟萍[1] 曼世超 LUO Yang-yi;LU Hui-guo;JIANG Juan-ping;MAN Shi-chao(College of ElectronicEngineering,Chengdu University of Information Technology,Chengdu 6100225,China)

机构地区：[1]成都信息工程大学电子工程学院,四川成都610225

出　　处：《成都信息工程大学学报》2018年第6期632-638,共7页Journal of Chengdu University of Information Technology

基　　金：四川省教育厅重点科技计划资助项目(14ZA0170)

摘　　要：随机森林是21世纪提出的基于分类树的算法,具有学习过程快速、运算速度快、稳定性好、预测精度高的优点。使用2017年8月雅安市泥巴山站点的电线覆冰数据训练一个用于预测覆冰现象是否出现的预测模型,并使用2017年11月-2018年1月的观冰数据对这一模型进行测试。测试结果表明该模型具有92.16%的查准率和88.26%的查全率。而使用2017年11月-2018年1月的电线覆冰数据训练得到的覆冰重量回归模型较线性回归模型有着更小的均方误差,随机森林回归模型的非线性特性能够更好地拟合覆冰重量。实验表明:随机森林算法既可以用于预测覆冰现象的出现,也可用于预测覆冰重量的变化;即随机森林算法在电线覆冰检测领域具有巨大潜力。Random forest is an algorithm proposed in the 21 st century based on classification tree.It has the advantages of fast learning process,fast operation speed,good stability and high prediction accuracy.Use wire icing datafrom Nibamountain site of Yaan city in August 2017 training a prediction modelused to predict whether icing phenomenon will be appeared,and use the icing data between November and January to test the model in this article.The test results show that the predictionmodel’sprecision ratio is 92.16% and its recall ratio is 88.26%.Use wire icing data between November and January train an ice weight regression model thathas a smaller mean square error rather than linear regression model.Because of the nonlinear characteristic,random forest regression model has a better ability in fitting the ice weight.The experiment shows that the random forest algorithm can be used to predict the occurrence of wire icing phenomenon,and it also can be used to predict the change of ice weight.In other words,the random forest algorithm has greatly potential ability in the field of wire icing detection.

关 键 词：随机森林 电线覆冰检测 查准率 查全率 均方误差 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程] TPL81[自动化与计算机技术—控制科学与工程]