Transformer Internal Fault Detection Based On Symbolic Dynamic

Transformer is the core electrical equipment of the electric power system,and it’s reliablity directly affect the running of electric power system.As a result,condition-based maintenace for transformer is essential.When implementing the condition-based maintenance,we should evaluate the state of transformer accurately so that operation teams can set out different maintenace strategies based on the state evaluation.And the internal fault degree is a important criterion which decide the transformer should be repaired or not.At the present time,transformer’s on-line monitoring methods include dissolved gas analysis,partial discharge detector,furfural analysis,infrared thermometry and so on,and all the methods above need complex sensors to support data.So this paper is aim to exploring a kind of relatively simple on-line monitoring method based on Symbolic Dynamic.The whole paper is divided into servral parts.The beginning of this paper shows the author’s studey on transformer’s simulation and classical transformer model.On the basis of former simulation model,we simulated the transformer internal fault and get the source current data under this condition.In the derivation process of the simulation model,we verified that Symbolic Dynamics is a feasible way to detect the transformer internal fault.Then we build mechanism of signal difference measurement based on the Symbolic Dynamic.We translated the current data to symbol sequence by phase space partition.We proposed a improved phase space parition method based on the concept of maximum entropy and difference partition method so that the symbol sequence can preserve the information of current signal as much as possible.Afterwards,this paper studied and used dictionary method and markov machine method respectively.The basic ideas was to transform symbol sequence into a pattern which has some inherent feature and use the difference of the feature to represent the signal’s deviation.We used different deviation index for dictionary method and markov machine method.The analysis of the experiment result is also presented in the paper.The experiment began with testing sine signal.The the result shows that two method mentioned above can detect the amplitude variations of signal.Then,simulation shows that when the inter-turn impedance becomes lower the method can effectively detect the difference between the normal and fault signal,which provides the basis of the condition-based maintenance.Moreover this paper also tested the Symbolic Dynamic method on the field test data of transformer internal fault,the result also verified the validity of the Symbolic Dynamic on the detection of transform internal fault.