Abstract: In this paper, using time Petri nets (TPNs) as specification models, system functionality and timeliness are first verified. Then a heuristic algorithm is proposed to optimize the system energy, which is driven by subtasks' energy-gradients and can be further simplified in the case that TPNs are composable. The experimental results show that the proposed methods can support the high-level low-power synthesis of real-time systems with results close to the optimum one, but with very low time complexities.