Research On Transformer Faiilt Diagnosis And Gondition Evaluation Technology Based On DSmT Information Fusion

The power transformer is responsible for the transmission,distribution and voltage conversion of electric energy in the power system.The operation status directly affects the overall safe operation of the entire power system.With the development of information fusion and artificial intelligence technology,the accurate and intelligent fault diagnosis and status assessment of transformers can not only provide guidance for the maintenance decisions,but also improve the operational reliability of transformers.Firstly,through analyzing and summarizing the state information of the transformer,the transformer fault diagnosis index system and the transformer state evaluation index system are constructed according to the requirements of the diagnosis and evaluation model under the condition of satisfying the rules of the index system.The state evaluation index system is divided into an indicator system based on component status level assessment and an indicator system based on typical defect risk level assessment.Aiming at the current situation that the shallow machine learning theory is not accurate in the transformer fault diagnosis and most diagnostic methods only refer to a single information feature,a transformer fault diagnosis method is proposed based on AdaBoost_RBF and DSmT(Dezert-Smarandache theory).Select the dissolved gas in the oil?test and gas production rate data to form the diagnostic parameter space which can reflect the transformer fault information.The RBF neural network algorithm is improved by using the AdaBoost algorithm.A parallel training unit is constructed with AdaBoost_RBF to construct the basic belief assignment(BBA)for the transformer fault recognition framework.Based on the idea of multi-source information fusion,the final diagnosis conclusion can be achieved by applying DSmT theory to fusing the BBA,which overcomes the limitations of D-S evidence theory that can not solve the fusion problem of high conflict evidences.Through the examples of 110 kV transformer,the result shows that the method has good practicability.Aiming at the randomness of state information and the fuzziness of evaluation results in transformer state evaluation,a transformer multi-level condition evaluation model is proposed based on cloud matter element algorithm and improved DSmT.The model is divided into transformer component condition level evaluation and transformer typical defect risk level evaluation.Based on the cloud matter element theory,the cloud matter element model of each test module and various typical defects of transformers are constructed.The AHP-entropy weight comprehensive weighting method is used to obtain the subjective and objective weights and improve the DSmT.The condition level membership degree of each test module of the transformer is used to obtain the condition level membership degree of the transformer components base on the improved DSmT.If the condition level of body is determined to be “abnormal” or “serious”,the risk level of defect type is evaluated based on the typical defect cloud matter element model which is constructed.The membership degree of risk level and risk score of the typical defect can be obtained after the DSmT fusion.Through the simulation analysis of the case,it is proved that the evaluation model has certain guiding significance for the maintenance decision of the transformer.