As the operating years of large-scale power transformers accumulate, their aging-related issues become increasingly prominent. To enhance the work efficiency of power grids in managing aging transformer fleets, this paper proposes a collaborative optimization method for preventive overhaul/retirement planning of transformer fleets considering aging failures and system reliability. Firstly, a novel aging failure model that considers the impact of preventive overhaul/retirement strategies is established based on the data-model hybrid-driven analysis technique of a transformer degradation trajectory. It can quantitatively capture the effects of implementing preventive overhaul/ retirement strategies at different times on transformer reliability by dynamically calibrating the aging threshold and expected lifetime parameters. Next, the unavailability calculation formula for transformer aging failure is derived using the integral discretization method. Then, an analytical calculation model for system reliability indices related to the unavailability parameters of the decision-making transformers is built based on the total probability formula. A reliability- constrained collaborative optimization model for coordinated overhaul/retirement strategies is established aiming to minimize the sum of costs of planned overhaul/retirement investment, fault disposition and reliability loss costs within the planning period. A hybrid approach combining simulated annealing (SA) and particle swarm optimization (PSO) is used for iteration analysis. Finally, case studies are conducted on the modified RTS-79 system to validate the effectiveness of the proposed method.