基于稳健估计的迭代型支持向量机及其在电站数据-中国电机工程学报.PDFVIP

第 31 卷 第 11 期 中 国 电 机 工 程 学 报 Vol.31 No.11 Apr.15, 2011 2011 年 4 月 15 日 Proceedings of the CSEE ©2011 Chin.Soc.for Elec.Eng. 113 0258-8013 (2011) 11-0113-06 TM 621 A 470 20 文章编号： 中图分类号： 文献标志码： 学科分类号： ⋅ 基于稳健估计的迭代型支持向量机及其在 电站数据检验中的应用 周卫庆，司风琪，乔宗良，周建新，徐治皋 ( 东南大学能源与环境学院，江苏省 南京市 210096) Iterative Support Vector Machine Based on Robust Estimation and Its Application in Data Validation in Power Plant ZHOU Weiqing, SI Fengqi, QIAO Zongliang, ZHOU Jianxin, XU Zhigao (School of Energy and Environment, Southeast University, Nanjing 210096, Jiangsu Province, China) ABSTRACT: A iterative support vector regression algorithm 归误差引起的残差污染，进而通过比较实测值与回归估计值 based on extended Huber estimation was proposed and applied 之间的残差来识别测量数据中的异常值。同时还给出了基于 to data validation and reconstruction. Extended Huber (eHuber) 该方法的数据检验及重构的流程与步骤。以某600 MW 机组 estimation of residual instead of sum of the least square of the 为例，对主蒸汽温度进行了算例分析，结果表明该方法能够 residual in the objective function of least squares support 有效识别主蒸汽温度测量值中异常值，并可给出可靠的重构 vector machine (LS-SVM) was used, and then an iterative 值，具有较好的稳健性能，而且迭代收敛较快，适合于工程 modelling algorithm was proposed. This modeling algorithm 上的在线应用。 can suppress residual contamination caused by regression error 关键词：电站；异常数据；稳健回归；支持向量机；数据检 of few abnormal data since considering the smooth of 验 regression curve entirety on the whole. So the outlier of 0 引言 measured data can be identified by comparing the regression value with the operational data. Then the framework of