The accurate life evaluation and operation state assessment of transformer is of great significance to the formulation of its maintenance strategy. In order to achieve the objective and scientific assessment of transformer life and state, this paper proposes a transformer life prediction and state assessment method based on multi-feature diagnostic parameters of adaptive fuzzy neural network (ANFIS). Firstly, the characteristic parameters affecting the life of the transformer are extracted, and these characteristic parameters are learned through an adaptive fuzzy neural network. The back propagation algorithm is adopted to solve the adaptive dynamic adjustment of the weights, and a life prediction model of the transformer is constructed; on this basis, a comprehensive health state assessment model of transformer is established by combining the dissolved gas in the oil. Through the research of experimental data, the model can accurately and effectively diagnose the transformer life and state, and has higher prediction accuracy and evaluation accuracy than the traditional method. It is a new and effective transformer state evaluation method.