Iterative Support Vector Machine Based on Robust Estimation and Its Application in Data Validation in Power Plant

A iterative support vector regression algorithm based on extended Huber estimation was proposed and applied to data validation and reconstruction.Extended Huber(eHuber) estimation of residual instead of sum of the least square of the residual in the objective function of least squares support vector machine(LS-SVM) was used,and then an iterative modelling algorithm was proposed.This modeling algorithm can suppress residual contamination caused by regression error of few abnormal data since considering the smooth of regression curve entirety on the whole.So the outlier of measured data can be identified by comparing the regression value with the operational data.Then the framework of data validation and reconstruction was presented.Main steam temperature in 600 MW thermal power plant was selected to be as an example,simulation results show that the method can distinguish the abnormal data effectively and give the reliable reconstruction value.When compared to other robust SVM,it revealed that the eHuber-SVM has fairly good robustness as well as less computing complex,it can be applied online data validation.