In order to predict the running state of power transformers

we firstly analyzed the influencing factors of the state during the operation of transformers

and selected the Gases concentration dissolved in insulation oil

operation condition

maintenance records and running time as the key influencing factors.Then

we used the fuzzy synthetic judgment method to evaluate the running status of power transformers

and established a state prediction model for power transformers based on the long short-term memory network with the fuzzy synthetic membership.The key parameters were trained to explore the prediction characteristics.The case studies show that the proposed method can effectively predict the running state of the power transformer.The accuracy of a week short-term prediction for power transformer state is 94.4%

and the accuracy of a month long-term prediction for power transformer state is 81.2%.The proposed prediction model can accurately reflect the reliability of the transformers.