滑动去趋势波动分析与近似熵在动力学结构突变检测中的性能比较  

Comparison of characteristics of moving detrended fluctuation analysis with that of approximate entropy method in detecting abrupt dynamic change

作　　者：何文平[1] 吴琼[2] 张文[3] 王启光[4] 张勇[1] 

机构地区：[1]国家气候中心,北京100081 [2]中国气象科学研究院,北京100081 [3]南京邮电大学通达学院,南京210003 [4]扬州大学物理科学与技术学院,扬州225002

出　　处：《物理学报》2009年第4期2862-2871,共10页Acta Physica Sinica

基　　金：国家重点基础研究发展计划(批准号:2004CB418303,2006CB400503);国家科技支撑计划(批准号:2007BAC03A01,2007BAC29B01)资助的课题~~

摘　　要：近似熵(ApEn)被认为是一种有效的动力学结构突变检测方法.将一种新的动力学结构检测方法——滑动去趋势波动分析(MDFA)与ApEn的检测结果进行了比较,检验了新方法的性能.结果表明,新方法的检测结果几乎不依赖于子序列的长度,而ApEn虽然能在一定程度上识别系统的动力学结构突变,但其检测结果依赖于子序列长度,且不能准确地检测出突变点的位置.因此,相对于ApEn方法而言,MDFA方法更适合于动力学结构突变检测,其优越性是显而易见的.Approximate entropy（ApEn）is well known to be an effective abrupt change detection method in dynamic structure.Based on this,we compare the performances between the moving detrended fluctuation analysis（MDFA）and ApEn in detecting abrupt dynamic change.The results show that the MDFA results almost do not depend on length of subseries,and while the ApEn could identify dynamic structure to some extent but still depends on the length of subseries.At the same time,there exists a huge drift for the ApEn results which means the actual time-instants of abrupt change don＇t coincide with the detected one.Therefore,compared with ApEn,MDFA is more suitable to be used to detect abrupt dynamic change,and the advantage of MDFA is obvious.

关 键 词：滑动去趋势波动分析 近似熵 动力学结构突变 

分 类 号：P46[天文地球—大气科学及气象学]