Spacecraft radiators can be designed through thermal analysis of a thermal model divided into nodes. The radiator is thus designed discretely in unit nodes; then, the radiator node distribution over the node division of a candidate radiator region indicates the radiator design as radiator node combinations. In particular, the radiator sizes and the topologies of shapes and locations are key design factors for efficient thermal design. These design factors can be considered, to some extent, at the typical resolution levels for the node division of the thermal model. This paper proposes two radiator design optimization methods that are based on node division of the thermal model; these methods find the optimal solutions of radiator node combinations in different ways. The first method uses an integrated optimization analysis that combines an optimization algorithm with thermal analysis; binary design variables are assigned to each node division to represent the radiator node combination and serve as the interface between two sets of software. The optimization problem is formulated as a multi-objective problem and improved through thermal design heuristics. The second method is a stepwise approach that adds a radiator node with the highest temperature sensitivity until the temperature limits are satisfied. A thermal model of a small spacecraft was developed to verify the proposed optimization methods, and test problems of various versions of the thermal model were appropriately defined. The numerical optimal solutions for the test problems using these methods showed good agreement with the analytic solutions. Therefore, the applicability and feasibility of the present methods for practical radiator design were confirmed.