In the thermal management of spacecraft, space thermal radiators play a vital role as heat sinks. A serial radiator with proven advantages in ground applications is proposed and analyzed for space applications. From the performance analysis, specific heat rejection (SHR) of serial radiator is found to be higher than parallel radiator by 80% for maximum diameter of the tube, 47% for maximum thickness of the fin, and 75% for maximum pitch of the tubes under consideration. Also, serial radiator requires four times higher pumping power than parallel radiator with geometric parameters and a maximum mass flowrate under consideration. In serial radiators, the cross conduction between the fins has a significant effect on its thermal performance. Thus, conjugate heat transfer simulations and optimization operations are to be performed iteratively to optimize the serial radiator, which is computationally costly. To reduce the computational time, artificial neural network (ANN) is trained using conjugate heat transfer simulations data and combined with the genetic algorithm (GA) to perform optimization. Taguchi’s orthogonal arrays provided the partial fraction of conjugate heat transfer simulations set to train the ANN. Taguchi-Neuro-Genetic approach, a process that combines the features of three powerful techniques in different optimization phases, is used to optimize both parallel and serial radiators. The optimization aims to obtain a configuration that provides the lowest mass and lowest pumping power requirement for given heat rejection. Optimization results show that the conventional parallel radiator is about 20% heavier and requires about 35% more pumping power than the proposed serial radiator.