Abstract: The transformer winding withstands the short-circuit force during transformer short-circuit. The transformer windings takes cumulative damage, which seriously affects the safe and stable operation of the power grid. However, previous researches have limitations such as insufficient extraction accuracy and inadequate generalization ability of prediction model when dealing with high-dimensional nonlinear features. Therefore, this paper develops an improved differential evolution algorithm for the extraction and prediction of cumulative damage characteristics of transformer winding short-circuit. In this paper, adaptive mutation strategy and elite reservation mechanism are integrated into the traditional differential evolution algorithm to improve the global search ability, and the improved algorithm is applied to the optimal extraction of damage features, which realizes the intelligent attenuation of key feature parameters, reduces feature maintenance, and improves the ability of feature expression. The experimental results show that the proposed method can effectively extract the essential characteristics of transformer winding cumulative damage, and greatly improve the prediction accuracy, which supports the safe and stable operation of the transformer.