Fault Diagnosis Method of Transformer Based on Rough Set and Support Vector Machine
NI Yuan-ping · Gao dianya jishu · 2008
A new power transformer fault diagnosis method based on the rough set theory(RST)and support vector machine(SVM)is presented for monitoring transformer latent faults and timely and accurately diagnosing faults.Firstly,a power transformer fault delamination diagnosis model was constructed.The rough set theory was applied to simplify expert knowledge,so that the minimal diagnostic rules could be obtained,and a transformer fault was roughly diagnosed by the rough set theory.Then,with exactly two-sorts classified function of support vector machine,fault diagnosis of power transformer was finally realized.Two intelligent arithmetics can be mutually repaired effectively in this method,which not only holds advantages of disposing imperfectness information competence,simple and speedy of rough set theory,also has exactly two-sort classify function of support vector machine.So the demerits of single intelligent arithmetic are compensated effectively,then shortcut and veracity of the power transformer fault diagnosis and extensive competence can be improved.Meanwhile fault swatch training time and diagnostic complexity are reduced by this method.Finally,the method was tested by many practical fault data of dissolved gas in oil of power transformer.Compared experimental results with those by the traditional three-ratio method and improved new IEC three-ratio method,the diagnosis accuracy is increased to 90%(higher),and the method has less fault tolerance for power transformer fault diagnosis.It is clearly shown that the method is valid and feasible and has better diagnostic accuracy.