Research On Fault Classification And Location Method Of Transformer Equipment Based On Sound Features

Power transformer substation fault detection technology,it is to point to by monitoring the transformer running state to test whether the power transformer substation is still normal work,if the failure happens,can timely alarm,convenient maintenance staff on the detection,but also can predict transformer over a period of time in the future work.With the development of industry,science and technology progress,people living standard unceasing enhancement,real-time on-line fault detection technology has gradually developed,power transformer substation fault detection scheme is proposed in this paper,through analysis to extract the transformer by the amplitude frequency characteristics of sound,and combined with the corresponding detection algorithm to achieve the purpose of transformer fault detection,can be used as an effective auxiliary transformer fault detection method.First,a sound collection model is established,and a multi-sensor sound acquisition system is constructed and designed to be used for sound collection and spatial sound source positioning.A central symmetric non-uniform microphone acquisition array is implemented.Microphone array can make full use of the voice signal characteristics of space and time,for jamming signal has strong anti-interference ability,can improve the quality of the collection to the goal of the voice signal,and can be used for space sound source localization.When the voice of the transformer is collected,the test system can judge the sound,and then the purpose of testing the transformer operation is achieved.Secondly,this paper proposes a power transformer amplitude frequency characteristics of sound data feature extraction methods,both can effectively extract the characteristics of sound,improve the fault identification accuracy,and can eliminate the interference signal of the sound samples,increase the accuracy of fault location.Application of two-dimensional principal component analysis(2 dpca)algorithm on the frequency spectrum characteristics of sound data dimension reduction processing,the main feature information extraction,using support vector machine(SVM)algorithm of voice signals are classified,in order to determine whether a power transformer is in a state of normal operation.Finally,the fault diagnosis of power transformer,this paper not only confined to diagnose the fault type,at the same time using the noise subspace and signal subspace orthogonal direction features,this paper proposes a weighted on the noise subspace of MUSIC algorithm,effectively solves the power transformer fault signal source DOA estimation problem.When power transformer fault occurs,can help the operation maintenance personnel to quickly determine the fault section,narrow your search,make emergency workers can work accurately and efficiently reach the point of failure.