基于随机森林算法的电线覆冰检测技术
|
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.
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.
引文
[1]陆佳政,蒋正龙,雷红才,等.湖南电网2008年冰灾事故分析[J].电力系统自动化,2008,32(11):16-19.
[2]牛生杰,周悦,贾然,等.电线积冰微物理机制初步研究:观测和模拟[J].中国科学:地球科学,2011(12):1812-1821.
[3]牛生杰,李蕊,吕晶晶,等.三种下垫面温度及结冰预报模型研究[J].地球物理学报,2011,54(4):909-917.
[4]刘雪静,牛生杰.两次高压电线积冰过程气象成因分析[J].气象科学,2016,36(2):230-235.
[5]MakkonenL.Modeling of ice accretion on wires[J].JClim ApplMeteor,1984,23(6):929-939.
[6]Finstad K J,Lozowski E P,Gates E M.A Computational Investigation of Water Droplet Trajectories[J].Journal of Atmospheric&Oceanic Technology,1958,5(1):160-170.
[7]廖玉芳,段丽洁.湖南电线覆冰厚度估算模型研究[J].大气科学学报,2010,33(4):395-400.
[8]郑振华,刘建生.遗传算法与BP神经网络相结合的输电线路覆冰厚度预测方法[J].电网与清洁能源,2014,30(4):27-30.
[9]黄宵宁,许家浩,杨成顺,等.基于数据驱动算法和LS-SVM的输电线路覆冰预测[J].电力系统自动化,2014,38(15):81-86.
[10]尹子任,苏小林.基于粒子群算法优化支持向量机的输电线路覆冰预测[J].电力学报,2014(1):6-9.
[11]Quinlan J R.C4.5:programs for machine learning[J].1993(1).
[12]Breiman L,Friedman J H,Olshen R,et al.Classification and Regression Trees[J].Encyclopedia of Ecology,1984,40(3):582-588.
[13]Breiman L.Random Forests[J].Machine Learning,2001,45(1):5-32.
[14]Liu F T,Ting K M,Zhou Z H.Isolation-Based A-nomaly Detection[J].Acm Transactions on Knowledge Discovery from Data,2012,6(1):1-39.
[2]牛生杰,周悦,贾然,等.电线积冰微物理机制初步研究:观测和模拟[J].中国科学:地球科学,2011(12):1812-1821.
[3]牛生杰,李蕊,吕晶晶,等.三种下垫面温度及结冰预报模型研究[J].地球物理学报,2011,54(4):909-917.
[4]刘雪静,牛生杰.两次高压电线积冰过程气象成因分析[J].气象科学,2016,36(2):230-235.
[5]MakkonenL.Modeling of ice accretion on wires[J].JClim ApplMeteor,1984,23(6):929-939.
[6]Finstad K J,Lozowski E P,Gates E M.A Computational Investigation of Water Droplet Trajectories[J].Journal of Atmospheric&Oceanic Technology,1958,5(1):160-170.
[7]廖玉芳,段丽洁.湖南电线覆冰厚度估算模型研究[J].大气科学学报,2010,33(4):395-400.
[8]郑振华,刘建生.遗传算法与BP神经网络相结合的输电线路覆冰厚度预测方法[J].电网与清洁能源,2014,30(4):27-30.
[9]黄宵宁,许家浩,杨成顺,等.基于数据驱动算法和LS-SVM的输电线路覆冰预测[J].电力系统自动化,2014,38(15):81-86.
[10]尹子任,苏小林.基于粒子群算法优化支持向量机的输电线路覆冰预测[J].电力学报,2014(1):6-9.
[11]Quinlan J R.C4.5:programs for machine learning[J].1993(1).
[12]Breiman L,Friedman J H,Olshen R,et al.Classification and Regression Trees[J].Encyclopedia of Ecology,1984,40(3):582-588.
[13]Breiman L.Random Forests[J].Machine Learning,2001,45(1):5-32.
[14]Liu F T,Ting K M,Zhou Z H.Isolation-Based A-nomaly Detection[J].Acm Transactions on Knowledge Discovery from Data,2012,6(1):1-39.