There is a problem of multiple characteristic indices and multiple fault types in power transformer condition evaluation. Thus a state evaluation method of a converter transformer is proposed based on a fuzzy comprehensive assessment model and improved D-S evidence theory fusion of the Pearson correlation coefficient. The transformer condition assessment system is built, and the basic probability of the state index is obtained using hierarchical analysis and an improved triangular trapezoidal membership function. The Pearson correlation coefficient is used to improve the D-S evidence theory in adopting different characteristic indices, and then the assessing result of the transformer is obtained from comprehensive evaluation. In the case analysis, the diagnostic accuracy of this method can reach 95.83% by comparing the condition assessment result and the actual operation condition of every transformer. This can validate the feasibility and rationality of the method. This method has a certain reference value for transformer state evaluation. This work is supported by the National Natural Science Foundation of China (No. 61901101).