Research On Fault Diagnosis Methods Of Transformer Based On Rough Sets Theory And Greedy Algorithms

Power transformer is one of the important equipments in power operation system.Along with increased voltage level,because the internal structure of the transformer is complicated and electric field distribution is asymmetrical,the accident rate rise,which will not only bring great loss to itself,but power security system will be impacted.Therefore detecting and diagnosing transformer faults effectively and accurate judgment of transformer potential fault and important hidden danger will become very important.One novel and practical fault diagnosis method for power transformer will be researched and the shortcomings of the existing transformer fault diagnosis methods are improved.Analysis system of the transformer fault diagnosis will be designed.Routine tests and trial operations will also be discussed.Firstly,the thesis introduces the request and the meaning of fault diagnosis of transformer.After compared with a few current main methods of fault diagnosis of transformer,the thesis brings forward the meaning of the analysis method of oil dissolution gas of transformer in the field of checking the inside fault of transformer.The current analysis method of the oil dissolution gas of transformer has a few defects:terrible independency,strong limitation,inconsiderate of different factors of transformer fault etc.Thus,the thesis presents a new transformer fault diagnosis method bases on the volume of oil dissolution gas of transformer.According to the uncertainty of the collection of oil dissolution gas of transformer,the rough sets theory is introduced into the thesis because the rough sets theory can deal with the inaccurate problems with remaining key information and processing data to get knowledge least expression.The thesis uses rough sets theory deal with a great deal of history data,get final reduction rule,then reduce the uncertainty of the collection of oil dissolution gas of transformer.Secondly,in the method of fit for the overall position of the transformer fault diagnosis,the discretization of continuous variables bases on the method of combination of rough sets and Boolean logic.The method doesn't rely on expert experience,but the discretization depends on the special data of special model,which has strong independency.Attributes reduction and value reduction use discernibility matrix algorithm to reduct,which can get the result of fit for the overall position,but what is found in the process of operation is that it can not get least reduction rule when the discernibility matrix algorithm is dealing with a NP-hard problem.So the thesis brings forward a new practical transformer fault diagnosis method,which builds up decision table base on all the percentage of oil dissolution gas of transformer and important ratio,and uses greedy algorithm do discretization of continuous variables,attributes reduction and value reduction on the decision tables.In the whole process of reduction,it uses uniform reduction rule,which reduces the calculating time and quantity.The thesis verifies the good result of the new method of transformer fault diagnosis by simulation experiment.Finally,the thesis points the problem of the fault diagnosis method of transformer in theory research and practical application,and the further research work is prospected.