Characteristics prediction of transformers based on long short term memory network

The transformer is an important equipment in power system, whose operational reliability of the transformer is the crucial factor to the system stability. the characteristic prediction is an useful direction for health assessment and fault diagnosis of transformers. In this paper, the characteristic prediction of transformers based on long short term memory network is investigated, and the comparison of the prediction effect for two input models is carried out. The results show that the prediction effects of two input models are almost same, and the prediction effect of the single input model is a little better. Particularly, for CH 4 and CO, the correlation coefficient of the multiple input model is a little higher than the single one. For CO 2 , the prediction has better adopt the single input model. This paper provides a technical support to the characteristics prediction of transformers.