用统计学习理论预测变压器油中溶解气体浓度.pdfVIP

003 11 9 11 13 An Prediction Model for Dissolved Gas in Transformer Oil Based on Statistical Learning T heory 王　鹏, 许　涛 ( 华北电力大学电力工程系, 北京10 06) L E- : , ( SVM) Ad- 0, ûf (x ) - y û E, aBoost , SVM L E(y ) = , ûf (x ) - y û- E, ûf (x ) - y û E , , Abstract This paper presents a new method for predicting f (x ) = Xx + b, 5 ( X, F, F) = dissolved gas in transform er oil based on stat istical learning p p ( ) . 1 theory SVM combined with the AdaBoost This met hod X + C( Fi + F) = mi n, takes full advantage of the SVM to solve the small sample i= 1 i= 1 problem. Hence the accuracy of the result estimation is im- F, F proved and a better generalization ability is guarant eed. T he actual cases are em ployed t o demonstrat e the validity of t he proposed approach. , : Key words transformer support vector machinegas-in- , oil concentrations TM 76 A , 0 p f (x ) = (Ai - Ai ) K (x i ,x ) + b, , i= 1 Ai , Ai, : , p , i i