In this paper, we propose a technique that uses thermal measurement results for improved accuracy in thermal simulation of electronic apparatuses. Because the modeling of the electronic components in such apparatuses has hitherto been very poor, the thermal simulation results cannot satisfy the required accuracy. To solve this problem, we first represent a component as a set of cubic blocks with equivalent thermal conductivity and contact thermal resistance values, and then identify these values by using the thermal measurement results for the component. We regard the identification of parameters as an optimization problem that involves minimizing the difference between the predicted and measured results. To solve the problem, we combine genetic algorithms and a thermal simulation tool. Our technique was successfully applied to the construction of an accurate thermal model, which we validated by using thermal measurement results.