Transformer Condition Detection And Evaluation Based On Vibration Analysis Method

Power transformer is the key position of electric energy conversion in power grid.For industrial manufacturing,inhabitants’ life and public order,its security and stability is meaningful.To prevent the sudden failure of the transformer and cause adverse effects,on-line testing and gauging of transformer condition has become a momentous topic.The vibration analysis method is to research and study the signal of the transformer body to achieve assessment of transformer mechanical stability.This supply for the discovery of potential mechanical failures in transformers.This paper analyzes the vibration of transformer body from the generation,propagation and synthesis of vibration.The relevance between winding vibration and current and core vibration and voltage are derived in detail.Analyze the correlation between the vibration of the body and winding and core according to the principle of vector synthesis.Secondly,this article adopts the methods of time-domain general statistical characteristics and probability distribution characteristics,Fourier transform,cepstrum and wavelet packet transform.Analyze the characteristic difference of the measured signal.Summarize the characteristic parameters that have a large degree of distinction and usable reflection of transformer status.According to the analysis of vibration mechanism and characteristics,a layered evaluation algorithm for transformer state is established.That is,first identify the operating conditions of the transformer,and then perform statistical analysis with the parameters under the similar cases to achieve layered assessment of the transformer status.Using the spectral clustering algorithm,the high-dimensional complex transformer working condition parameter feature vector is mapped into a low-dimensional vector,the cluster analysis of the transformer working condition is efficiently and accurately realized,and the typical working condition library of the transformer is established.Using the mapping matrix coefficients to improve the discriminant formula,the transformer working condition can be accurately identified.The non-a priori frequency concentration index and a priori vibration stability index are summarized,and a transformer state evaluation model that meets the requirements of universality and accuracy is realized.Use T-square statistics to analyze the state evaluation characteristic indicators and determine the control limit.Finally,the effectiveness of the algorithm is verified by the laboratory platform,and,a layered evaluation algorithm procedure for transformer state is introduced.The transformer condition detecting and assessing system is designed.In terms of hardware,the circuit includes sensor conditioning,zero crossing comparison,current supply and communication.In terms of software,the transplantation of Free RTOS,data acquisition and handling,state evaluation algorithm,letter of agreement,HMI and background data management software are accomplished.Through the construction and test of the software and hardware of the system,the feasibility of the algorithm is verified.