Abstract: DVS(dynamic voltage scaling) and cache partition are two technologies aimed at reducing embedded systems' energy consumption. The combination of the two techniques will not only make a better energy saving, but also have other advantages of cache partition. The key issue is how to determine the voltage and cache size for each task to minimize the energy consumption. In real-time embedded systems, on this issue, a polynomial time complexity algorithm is proposed. Experiments compared the energy consumption of multiple test sets under different algorithms, and results show that the algorithm can effectively reduce the energy consumption under the real-time constraint.