There is a problem that the dynamic change of real-time condition information of the components in a maintenance time window is rarely considered in existing research on wind farm group maintenance optimization. Thus this paper proposes a wind farm group maintenance strategy that considers the dynamic update of the remaining life prediction information. First, the residual life prediction results of each component are obtained using real-time condition information. Based on these results, the minimum average maintenance cost is optimized, and the optimal maintenance time window of single component is constructed. Secondly, considering the structural correlation of wind turbine components and the inventory constraints of spare parts, a group maintenance model of the wind farm is established to save the maximum maintenance cost, and a genetic algorithm is used to optimize the group maintenance strategy. Finally, a rolling time window model is used to update the remaining life prediction information of wind turbine components in real time and dynamically adjust the original maintenance plan. The analysis of an actual wind farm case shows that the proposed strategy can update the wind farm maintenance plan in real time, realize the dynamic optimization of the maintenance plan, and help to reduce the maintenance cost.