A combined approach of analytical and numerical methods is proposed for the optimal arrangement of n heat sources to achieve the desired temperatures at m locations. From the characteristics of Green's function under a point-source approximation, it is shown that the optimal heat-source intensities can be determined non iteratively by executing n + 1 or m + 1 numerical simulations of heat conduction equations. Using the results of m + 1 numerical simulations, the optimal heat-source locations can also be obtained by the conventional gradient-based optimization strategy without additional numerical simulations. Moreover, these numerical simulations are independent of the existence of heat sources and of the desired temperatures. Therefore further numerical simulations are not required even if the number of heat sources and the desired temperatures are changed.