基于机器学习的雷电预报研究  

Study on lightning prediction based on machine learning

作　　者：崔月生 胡曦 Cui Yuesheng;Hu Xi(Beijing Lightning Devices Security Test Center,Beijing 100089,China)

机构地区：[1]北京市避雷装置安全检测中心

出　　处：《国外电子测量技术》2019年第6期23-27,共5页Foreign Electronic Measurement Technology

摘　　要：高准确率的雷电预报,有助于降低雷电带来的灾害,从而减少雷电造成的损失,所以如何提高雷电预报准确率具有重要的现实意义。为了提高支持向量机(SVM)算法的分类效果,引入了灰狼优化算法(GWO),利用GWO算法的全局优化能力优化SVM的c和σ,从而提高SVM分类的准确性。由于采集的雷电数据属性较多,采用主成分分析(PCA)方法对属性进行约简,获得能够反映雷电情况的主要影响因子,作为GWO-SVM的输入数据,GWO-SVM的输出为雷电发生情况。最后建立了雷电预报仿真实验,实验对比结果现实,在相同的实验参数及实验数据情况下,GWO-SVM方法相比于传统的其他3种算法具有更高的雷电预报准确率;相比于前人所作研究,所提方法也具有更高的雷电预报准确率;验证了所提雷电预报方法的可靠性。Lightning prediction with high accuracy can help to reduce the disaster and the the loss caused by lightning. Therefore,how to improve the accuracy of lightning prediction has important practical significance. In order to improve the classification effect of SVM algorithm, grey wolf optimization (GWO) algorithm is introduced. The global optimization ability of GWO algorithm was used to optimize c and σ of the SVM, so as to improve the accuracy of SVM classification. Since there are many lightning data attributes collected, the PCA method is adopted to reduce the attributes, and the main influencing factors that can reflect the lightning conditions are obtained, which are used as the input data of GWO-SVM, and the output of GWO-SVM is the lightning occurrence. Finally, the simulation experiment of lightning prediction is established. The comparison results show that the GWO-SVM method has higher lightning prediction accuracy than the other three traditional algorithms under the same experimental parameters and data. Compared with previous researches, the method proposed in this paper also has higher accuracy of lightning prediction. The reliability of the lightning prediction method is verified.

关 键 词：支持向量机 灰狼算法 雷电预报 主成分分析 

分 类 号：P427.32[天文地球—大气科学及气象学] TN0[电子电信—物理电子学]