The power transformer

as an important part of the entire power system

undertakes the work of powertransmission and distribution in the power grid

assures the effective conversion of voltage and therefore plays a vitalrole in power service. The operational variation of the transformer can be mastered by analyzing the feature of charac-teristic parameters at different types of operation conditions

the early detection on the fault of transformer is achievedand corresponding preventive measures are made so to avoid the occurrence of the fault. In this paper

a kind of opera-tion state analysis method of transformer based on deep belief network is proposed. The un-coded characteristic ratioof dissolved gases in the oil is calculated and input into the deep belief network so to explore the relationship betweenthe data information and the operation condition of transformer as well as judge the operation condition of the trans-former. It is shown by the experimental result that the method in this paper has higher analysis accuracy.