Robust Statistics - 2nd Edition
鲁棒统计，现代统计方法， Robust Statistics第二版，学习现代统计方法R○ BUST STAT|STCSSecond editionPeter j, huberProfessor of Statistics, retiredKlosters SwitzerlandEⅣ ezio m. RonchettiProfessor of StatisticsUniversity of Geneva, SwitzerlandWILEYA JOHn WileY SONS INC. PUBliCAtIONCopyrightc 2009 by John Wiley Sons, Inc. All rights reservedPublished by John Wiley sons, Inc, Hoboken, New JerseyPublished simultaneously in CanadaNo part of this publication may be reproduced, stored in a retrieval system, or transmitted in any form orby any means, electronic, mechanical, photocopying, recording, scanning, or otherwise, except aspermitted under Section 107 or 108 of the 1976 United States Copyright Act, without either the priorwritten permission of the Publisher, or authorization through payment of the appropriate per-copy fee tothe Copyright Clearance Center, Inc, 222 Rosewood Drive, Danvers, MA 01923, (978)750-8400, fax978)750-4470,oronthewebatwww.copyrigom. requests to the publisher for permission shouldbe addressed to the permissions department John Wiley sons, Inc., 11 1 River Street, Hoboken, NJ07030,(201)748-6011,fax(201)748-6008,oronlineathttp:/www.wileycom/go/permissionLimit of Liability /Disclaimer of Warranty: While the publisher and author have used their best efforts inpreparing this book, they make no representations or warranties with respect to the accuracy orcompleteness of the contents of this book and specifically disclaim any implied warranties ofmerchantability or fitness for a particular purpose. No warranty may be created or extended by salesrepresentatives or written sales materials. The advice and strategies contained herein may not be suitablefor your situation. You should consult with a professional where appropriate. Neither the publisher norauthor shall be liable for any loss of profit or any other commercial damages, including but not limitedto special, incidental, consequential, or other damagesFor general information on our other products and services or for technical support, please contact ourCustomer Care Department within the United States at(800)762-2974, outside the United States at(317)572-3993 or fax(317)572-4002.Wiley also publishes its books in a variety of electronic formats. Some content that appears in print maynot be available in electronic format. For information about wiley products, visit our web site atwww.wileycomLibrary of Congress Cataloging-in-Publication Data:Huber Peter JRobust statistics, second edition/ Peter J. Huber, Elvezio ronchettip. cnIncludes bibliographical references and indeISBN978-0-470-12990-6( cloth)1. Robust statistics. I. Ronchetti. elvezio. II. TitleQA276.H7852009519.5-dc222008033283Printed in the United States of america10987654321To the memory o1John w. tukeyThis Page Intentionally Left BlankCONTENTSPrefacePreface to first editionGeneralities1 Why robust Procedures1. 2 What Should a robust procedure achieve?1.2.1 Robust. Nonparametric and Distribution-Free1.2.2 Adaptive procedures1.2.3 Resistant Procedures1.2. 4 Robustness versus Diagnostics1.2.5 Breakdown point1.3 Qualitative Robustness567888911. 4 Quantitative Robustness1.5 Infinitesimal Aspects141.6 Optimal Robustness171.7 Performance Comparisons18CONTENTS1.8 Computation of robust estimates181.9 Limitations to Robustness Theory202 The Weak Topology and its Metrization23eneral remarks232.2 The Weak Topology232.3 Levy and prohorov metrics272.4 The bounded Lipschitz metric322.5 Frechet and Gateaux derivatives366 Hampels Theorem413 The Basic Types of Estimates453. 1 General Remarks453.2 Maximum Likelihood Type Estimates(M-Estimates)3.2.1 Influence Function of m-estimates73.2.2 Asymptotic Properties of M-Estimates483.2.3 Quantitative and Qualitative Robustness of MEstimates3.3 Linear Combinations of Order Statistics(L-Estimates)3.3.1 Influence Function of -Estimates3.3.2 Quantitative and Qualitative robustness of l-Estimates 593. 4 Estimates Derived from Rank Tests(R-estimates3.4.1 Influence Function of R-Estimates623.4.2 Quantitative and Qualitative robustness of R-Estimates 643.5 Asymptotically Efficient M-, L,and R-Estimates674 Asymptotic Minimax Theory for Estimating Location4.1 General remarks4.2 Minimax bias4.3 Minimax Variance: Preliminaries744. 4 Distributions minimizing fisher Information764.5 Determination of Fo by Variational Methods814.6 Asymptotically Minimax M-Estimates914.7 On the minimax Property for L-and R-estimates954.8 Redescending m-estimates74.9 Questions of Asymmetric Contamination101CONTENTSScale Estimates1055.1 General remarks1055.2 M-Estimates of scale1075.3 L-Estimates of scale5.4 R-Estimates of Scale1125.5 Asymptotically efficient Scale estimates1145.6 Distributions Minimizing fisher Information for Scale5.7 Minimax Properties116 Multiparameter Problemsin Particular Joint Estimationof Location and scale1256. 1 General remarks1256.2 Consistency of M-Estimates1266.3 Asymptotic Normality of M-Estimates1306. 4 Simultaneous m-Estimates of Location and scale1336.5 M-Estimates with Preliminary Estimates of Scale1376.6 Quantitative robustness of Joint Estimates of Location and Scale 1396.7 The Computation of M-Estimates of Scale14368Studentizing1457 Regression1497. 1 General remarks1497. 2 The Classical Linear Least Squares Case1547. 2.1 Residuals and Outliers1587.3 Robustizing the Least Squares Approach1607.4 Asymptotics of robust regression Estimates163741 The Cases hp2→0 and hp→07.5 Conjectures and Empirical Results1687.5.1 Symmetric Error Distributions1687.5.2 The Question of Bias1687.6 Asymptotic Covariances and Their estimation1707. 7 Concomitant Scale estimates1727.8 Computation of Regression M-Estimates1757.8.1 The Scale Step1767.8.2 The Location Step with Modified residuals1787.8.3 The Location Step with Modified Weights179CONTENTS7.9 The Fixed Carrier Case: What Size hi?1867. 10 Analysis of Variance1907. 11 LI-estimates and Median polish1937. 12 Other Approaches to Robust Regression1958 Robust Covariance and Correlation Matrices1998. 1 General remarks8.2 Estimation of Matrix Elements Through robust Variances2038.3 Estimation of Matrix Elements Through robust Correlation2048.4 An Affinely equivariant approach2108.5 Estimates Determined by Implicit Equations2128.6 Existence and Uniqueness of Solutions2148.6. 1 The Scatter estimate v2148.6.2 The Location estimate t2198.6.3 Joint Estimation of t and y2208.7 Influence Functions and Qualitative robustness2208.8 Consistency and asymptotic normality2238.9 Breakdown Point48.10 Least informative distributions2258.1058. 10.2 Covariance2278.11 Some Notes on Computation2339 Robustness of Design2399.1 General remarks2399.2 Minimax Global Fit9.3 Minimax Slope24610 Exact Finite Sample Results24910.1 General Remarks24910.2 Lower and Upper Probabilities and Capacities25010.2.1 2-Monotone and 2-Alternating Capacities25510.2.2 Monotone and Alternating Capacities of Infinite Order 25810.3 Robust Tests25910.3. 1 Particular Cases26510.4 Sequential Tests267
- 2020-12-03下载
- 积分：1
PSCAD_EMTDC与Matlab接口
PSCAD_EMTDC与Matlab接口，用PsCAD建立接口模型．启动Matlab数据引擎调用M文件，实现接口模型的参数设置。第I了期杨使,等:FCAD/ EMTDC与Maah接口研究500∠0°kV下的样本,利用接口软件所获得的数据经过计算小500∠0°kY500∠30°kvc波能量熵后,将其随机分为2个部分,一部分为训练--50 km样本,另一部分为测试样本,这样极大减轻了工作量。100 k2764k训练样本与训练的目标向量作为神经网络输入数1μF据进行网络训练,测试样本输入训练后的网络,将输壬出与期望输出进行比较,进而得到暂态识别的结果。图9500kV输电线路模型示意图利用接口软件随机产生不同工况下的单相接地ig. 9 Model of 500 kv transmission line短路和开关操作数据共1000组采用3层BP神经型,本文模型的采样频率为40kHz线路模型采用网络对2种暂态信号进行分类,取用其中的200组频率相关模型。利用文献[2所述的小波能量熵测进行网络训练,其余800组用来测试。度来识别开关操作与单相接地故障。网络设计步骤如下识别各暂态信号,采用小波能量熵提取暂态信a,构造特征向量,直接选取各暂态信号在1~16号特征,然后将其作为神经网络的输人来进行识尺度上的小波能量熵作为特征向量;别。识别过程如图10所示。b.为了便于分析,归一化处理小波能量熵电压/电流信号「训练c设计输人层神经元共16个、隐含层神经元标向量小波变换」[训练样本P-网络训练33个输出层神经元1个;d设计网络的训练函数为 trainlm,训练误差小波时频嫡和小波L构造特测试样本训练后0001。熵权的征提取征向量「识别结果下的网络通过训练和测试,利用接口与3层BP神经网络图10暂态信号识别过程综合应用,成功地实现了暂态信号的分类,其中单相Fig 10 Process of transient signal recognition短路故障的误判率为5%,开关操作的误判率为0。人工神经网络:6的训练需要大量不同工况·表1给出了部分数据及其识别结果。表1小波能量熵神经网络识别结果举例Tab. 1 An example of wavelet energy entropy nn detection类别输人数据(归一化后)期望输出实际输出测试结果03121018660.312302709021930.22410,23080.23640%881正确开关0.24120.24580.2499025370257202604026340.2663操作0.25270.254202522025189.25190.25170251402508024650.2469024710.24800.24880.2501024880247410.9622正确0.325501410019560187502064022l8023430.2441单相025250.260102662027150.2762028040.28430287000178正确短路0387701584027200.186502023021510224402204005364错误0226902322024280253602626027130278202848注:在误差允许情况下,认为大于5为1小于.5为04接口软件运用的优点对梭型进行参数]运行程序]「得到1组短路电设置短路电阴为进行仿真阻为0时的数据通过对以上2个仿真实例的分析,对所研究的图11传统软件获取一组数据的示意图接口软件的运用有了更深入的了解。应用传统的仿Fig 11 Data group access of traditional software真软件获取数据,对于每一种工况的实验都需要对1组数据的示意图,此时短路电阻为0。如果要完成模型的参数以及仿真的条件进行重新设置,这样,做述10种不同工况的仿真就需要等待短路电阻为个大数据量的仿真实验,大部分的时间都将消耗0的仿真数据获取以后,修改模型参数,再进行实在重复的点击工作和等待上,工作效率不高;而这里验即要顺次进行图11的流程10次,才能达到最终所研究的接口软件与传统的仿真软件相比,最大的的目的。如果要获取1000组数据,那么就要顺次重优点就是可以一次获取大量不同工况下的数据,对复图1l的流程1000次,工作量之大,耗费时间之不同的工况进行伤真实验,只需要编写相应的M文长可想而知。传统仿真软件的这种缺点也迫切的需件,数据将实现自动获取极大解放了人的劳动。获要研究一种能实现数据自动获取的新型软件。取文中用于BP神经网络算法的1000组实验数据,图12是这里所研究的接口软件获取数据的示使用所研究的接口软件将带来很大的方便,下面以发意图。可知,对于单相短路故障时电阻由0~900g生单相短路故障时短路电阻分别为0、100、200、变化的仿真,只要编写相应的M文件,使电阻按照300400,500600、700、8009009的10种工况为需要变化,就可以实现数据的自动获取,在M文件例来说明应用本文接口软件的优点中,还可以设置电阻为0不变时,在不同的时刻发生图11是利用传统的仿真软件获取一种工况下故障,这样,通过设置,就可以实现仿真一次获取电力自动化设各第27卷电阻为0的数据and its application[J]. Electric Power Automation Equipment编写M文件运行程序进行仿分段电阻为10的数据2006,26(11):67-70使电阻按照真得到10种不[7]朱瑜,梁旭,闵勇,基于 PSCAD/ EMTDC的高压直流输电线路保合要变化同电阻时的数据」读出护仿真研究].现代电力2006,23(2):35-38电阻为909的数据ZHU Yu, LIANG Xu, MIN Yong Simulation of line protection of图12接口软件获取数据示意图HVDC transmission based on PSCAD/EMTDC[J ] ModernFig 12 DataElectric Power, 2006, 23(2): 35-388]张志酒.精通 Matlab65版[M].北京:北京航空航天大学出版个系统数据文件,它包含了所需要的不同工况下的杜,2003所有数据。接口软件由于要调用 Matlab的M文件[9韩笑,徐曦陈卓平.基干 Matlab与VB数据交换的继电保护仿而占用了更多的CPU时间,但由于仿真的过程中不真[J电力自动化设备,2006,26(5)92-95需要对模型参数作任何修改,编写适合的M文件HAN XiaO, XU Xi, CHEN Zhuo- ping. Protection simulation后,程序自动运行,数据文件自动存储,研究人员可based on data exchange between Matlab and VB[J].Electric利用程序运行的时间去做其他研究工作,这样就不Power Automation Equipment, 2006, 26(5): 2-95会像应用传统的仿真软件那样,把时间浪费在点击10许允之刘吴冯字等.Mulb在电力系统仿真实验中的应用[丁.实验技术与昏理,2007,24(1):103-105,125和等待上,大幅提高了研究人员的工作效率。XU Yun-zhi, U Hao, FENG Yu,ct al. The application of the5结论Matlab in the power system simulation experiments[J]. Experimental Technology and Management, 2007, 24(1): 103-105, 125.对学习与研究传统的继电保护算法以及新型的11.钟2,赵华,BCAD围MmDC程序与M言接口的研究电力智能算法而言,能方便、快捷地产生多种工况数据的ZHONG Bo, ZHAO Hua- jun. Study on PSCAD/EMTDC仿真软件是至关重要的。这里所研究的接口软件能Matlab interface[J]. Guangdong Electric Power, 2005, 18(8):1-3减少仿真时间,提高仿真效率,有助于研究者更好地12】何正友陈小勤.基于多尺度能量统计和小波能量嫡测度的电研究电力系统的规律。力暂态信号识别方法[J].中国电机工程学报,2006,26(10)这里所述的接口软件能对不同工况的故障进行仿真与分析。它能一次产生数以万计的不同工况的HE Zheng-you, CHEN Xiao-qin. A study of electric数据。此软件具有较好的解耦性:对于不同的电力system transient signala identification method based on模型以及不同的分析算法,都不需要进行大的改动,scales energy statistic and wavelet energy entropy[]. Proceedinge of the CSEE, 2006, 26(10>: 33-39甚至不必修改。[I3]李洪,王晟.基于小波包和神经网络的电力输电线故障诊断研小波能量熵与BP神经网络在暂态识别上确实究[].数据采集与处理,2004(4);16有较好的性能,但也存在误判的情况。此算法仍然L Hong, WANG Sheng. Fauit diagmosis in power transmission需要研究者的进一步分析研究。line based on wavelet packets and neural network approach[J]. Jourmal of Data Acquisition Processing, 2004(4): 1-6.参岩文献[14]杜选高明峰.人工神经网络在数字识别中的应用].计算机系统应用,2007(2):2127]吴天明,谢小竹,彭彬,等. Matlab电力系统设计与仿真!M].北京:国防工业出版社,2004[2]李广觊,李庚银.电力系统仿真软件综述电气鬼于教学学报,2005,27(3):61-65Applications, 2007(2): 21.2LI Guang-kai, LI Geng-yin. The summary of power system simu15]张亚軍,刘志刚,张大渡.一种基于多神经网络的组合负荷预lation software[J]. Journal of Electrieal Electronic Engineering测模型[J,电网技术,2006,30(21)21-25Education42005,27(3):61-65ZHANG Ya-jun, LIU Zhi-gang, ZHANG Da-bo. A combination[3]KEZUNOVIC M, CHEM Q Novel approach for interactive protectionload forecasting model based on multinetworks[JIsystem simulation [J]. IEEE Trans on Power Delivery, 1997Power System Technology, 2006, 30( 21>: 21-2512(2):668674矗任编辑:李玲)[4] ZHANG Nan, KEZUNOVIC M. Implementing an advanced simulalion tool for comprehensive fault analysia[J]. IEEE on Trans作者简介mission and distribution conference and exhibition .2055.15杨健牟(1983-),女,辽宁凌源人,硕士研究生,主要研究(18):1-6.方向为电力系统继电保护(E·mai: yangjlanwei0910@163[5]林良真,叶林,电磁暂态分析软件包 PSCAD/EMTDC[J.电网技术,200,24(1):65-65麦瑞坤(1980-),男,广东东莞人,博士研究生,主要研究LiN Lipng-zhen, YE Lin. An introduction to PSCAD/EMTDCEJIPower System Technology, 2000,24(1):65-66领战为电力系統故障诊断中信号处理和信息理论的应用、新6]肖异,尹项根,张哲等 PSCAD/EMTDC程序与继电保护仿真模型线路保护理接口技术及应用[J].电力自动化设备,2006,26(11):67-70何正友(1970-),男,四川自貢人,教授,博士研究生导师XAOY, YIN Xiang-g=n, ZHANG Zhe et al. Interface technique主要从事馆号处理和信息理论在电力系統故障诊嘶中的应between PSCAD/ EMTDC and relay protection simulation model用、新型继电保护原理、配电网自动化等方向的研究工作。第27卷第11期电力动化设备Vol27 No 112007年11月Electric Power Automation EyuipmentNy.2007种新的适合分布安装的消弧线圈唐轶,陈庆(中国矿业大学信电学院,江苏徐州221008)摘要;可自恢复性单相接地故障点的电弧是否能自然熜灭的决定因素是接地故障残流的大小。以降低单相接地故障点的残流为出发点,从理论上分析了谐振接地系统残流产生的原因:消弧线自动跟踪补偿只能有效地降低零序回路的无功电流,不能降低零序回路的有功电流。通过仿真计算得出结论:消弧线圈分布安装是降低谐振接地系统接地故障点残流有功分量的有效方法。针对我因6kⅤ和10k中压配电网一般均为Δ接线,无辅助中性点供消孤线图接入的特点,设计了一种新颖的消弧线图。该消弧线图为三相五柱电抗器结构,通过调节两边柱的气隙大小改变补偿电流的大小,结枘简单,适合于分布安装。实验室试验证明其补偿电感线性度好、补偿效果好。关键词:消弧线團;单相接地故障;中性点接地中图分类号;TM55文献标识码:A文章编号:1006-6047(2007)11-0087-04地故障电弧自行熄灭、故障自恢复为原则的。因此0引言消弧线圈的安装、运行应该以使接地故障点的残流配电网故障的80%左右为单相接地故障1。尽量小为目标。单相接地故障中的绝大多数为可自恢复的故障,尤不管是城市电网还是农村电网,用电负荷都在急其是自然条件差(台风、雷电频繁)的架空线电网需剧增加,电网的结构及规樸在不断扩大;城镇电网的要分断电路处理的永久性单相接地故障更是极少改造中,电缆网络正在逐渐取代架空线路;过去采用数。因此,从提高供电可靠性考虑,我国中压配电单电源的辐射式供电或树状供电方式,已不能满足用网绝大多数采用小电流接地方式。小电流接地电网电负荷增长的要求,而需要采用网孔形或环形等供电中,单相接地故障相当大一部分为可自恢复的故障。方式;这些因素都使其单相接地故障电流急剧增加可自恢复性单相接地故障点的电弧是否能自然熄灭单体大容量自动跟踪补偿消孤线圈被局。当的决定因素是接地故座残流的大小:残流小,有利于消弧线图的单体容量不能满足补偿电网接地电流的电弧过零时媳灭;残流小,电弧对介质绝缘的破坏程要求时在同一电网安装2台或多台自动跟踪补偿度低有利于故障点绝缘介质的恢复使电弧不易重弧线圈的也有之。实际上,消弧线圈只能减少接燃:残流小,也有利于降低故障相恢复电压的初始速地故障电流的无功分量,即脱谐度只是单相接地残度,使电弧不易重燃。小电流接地方式是以单相接流中无功分量大小的决定因素。即使采用自动跟踪补偿的方法来实现理想调谐,使接地电流中的无功收稿日期:2006-11-16;修回日期:2007-03-30分量几乎为零后,零序回路的有功损耗电流仍然不Interface between PSCAD/EMTDC and MatlabYANG Jian-wei, MAI Rui-kun, HE Zheng-youof elng, Southwest Jiaotong University, Chengdu 61003Abstract. Theen PsCad emtdc and matlaTo make ththe electromagnetic transient analysis program PSCAD/EMTDC and the math model software packageMatlab, the interface model is built using PSCAD and its parameters are set by calling the M filesing the data engine of Matlab. Massive data under different conditions could be accessed via thisinterface once it runs. An application example of power transmission line is analyzed. Data got viathe interface are processed in segments and sent to BP neural network to detect single -phaserounding fault from switch operations. Simulation results point out that this interface softwarefacilitates the acquisition of massive dataThe project is supported by National Natural Science Foundation of Ching(50407009)and DistinguishedScholars Fund of Sichuan Province(06ZQ026-012)Key words: PSCAD/EMTDC; Matlab; BP neural networkPSCAD/ EMTDC与Mat1ab接口研究旧WANFANG DATA文献链接作者:杨健维,麦瑞坤,何正友, YANG Jian-wei, MAI Rui-kun, HE Zheng-you作者单位:西南交通大学,电气工程学院,四川,成都,610031刊名:电力自动化设备 TICEIPKU英文刊名:ELECTRIC POWER AUTOMATION EQUIPMENT年,卷(期)2007,27(11)被引用次数1次参考文献(15条1.KEZUNOVIC M; CHEM Q Novel approach for interactive protection system simulation 1997(02)2.李广凯;李庚银电力系统仿真软件综述[期刊论文]电气电子教学学报2005(03)3.吴天明;谢小竹;彭彬 Matlab电力系统设计与仿真2004.韩笑;徐曦;陈卓平基于 Matlab与ⅦB数据交换的继电保护仿真[期刊论文]电力自动化设备2006(05)5.张志涌精通 Matlab6.5版20036.朱瑜;梁旭;闵勇基于 PSCAD/ EMTDC的高压直流输电线路保护仿真研究[期刊论文]现代电力2006(02)7.张亚军;刘志刚;张大波一种基于多神经网络的组合负荷预测模型[期刊论文]电网技术2006(21)8.杜选;高明峰人工神经网络在数字识别中的应用[期刊论文]计算机系统应用2007(02)9.李洪;王晟基于小波包和神经网络的电力输电线故障诊断硏究[期刊论文]数据采集与处理2004(04)10.何正友;陈小勤基于多尺度能量统计和小波能量熵测度的电力暂态信号识别方法[期刊论文]中国电机工程学报2006(10)11.钟波;赵华军 PSCAD/EMTDC程序与 Matlab语言接口的研究[期刊论文]广东电力2005(08)12.许允之;刘昊;冯宇 Matlab在电力系统仿真实验中的应用[期刊论文]实验技术与管理2007(01)13.肖异;尹项根;张哲 PSCAD/ EMTDO程序与继电保护仿真模型接口技术及应用[期刊论文]电力自动化设备2006(11)14.林良真;叶林电磁暂态分析软件包 PSCAD/EMTDC[期刊论文]电网技术2000(01)15. ZHANG Nan; KEZUNOVIC M Implementing an advanced simulation tool for comprehensive fault analysis2005(18)引证文献(1条)王朕.朱琳.温渤婴基于 PSCAD的继电保护电压电流发生器的硏制[期刊论文]电力自动化设备2010(8)本文链接http://d.g.wanfangdata.com.cn/periodiCaldlzdhsb200711021.aspx
- 2021-05-06下载
- 积分：1
|
联系客服（QQ：3324461878）
