In this paper, we propose an optimization methodology to find the optimal topology for the data communication network in distributed frequency control of power system. In order to implement a distributed cooperative control scheme, local controllers often share their data over a data communication network. Structure of this network has a major role in determining stability of secondary cooperative frequency control of a microgrid; and thus the structure can be optimized to have the best performance. Although distributed control signals may be delayed or dropped during their transmission, confident margin of the stability can reduce side effects of these inherent problems and makes power system more reliable. In this situation, the challenge is to find the best topology of data communication network giving the highest margin of the system stability. In this paper we define the problem of finding the best topology as an optimization problem. An eigenvalue perturbation analysis approach is used to approximate sensitivity of the stability performance of secondary cooperative control to adding/removing data communication links. Then, the optimization problem is solved using a simulated annealing optimization strategy. Our numerical simulations on sample networks show that the proposed rewiring-based optimization can successfully find the network structure with (near)-optimal stabilizability performance.