Abstract:Traditional transformer state detection mostly deals with data obtained from single sensor,ignoring the potential connection between different sensors,resulting in simultaneous interpreting.In order to avoid information island and efficiently use multi-source information,this paper uses the difference and complementarity of multi-source information to make up for the defect of detection based on single information,and detects multi-level,multi-faceted and multi-level information from different sources and different modes of multi-sensor information to obtain the state and feature estimation of power transformer fault,and then realizes the fault detection of power transformer.In order to improve the accuracy of the sensing information,a multi-dimensional state parameter information processing circuit,such as band-pass filter,energy level amplification and choroid detection,is designed.Then,the set pair analysis theory is introduced into the uncertainty of transformer health state evaluation,and the transformer health state evaluation model based on information fusion is established by using D-S evidence theory fusion,which improves the system overall on-line detection level and fault diagnosis ability.Finally,the multi-source information fusion model is used to verify the feasibility and superiority of the scheme.