Abstract: Different task scheduling algorithms assign tasks to different processors and result in different energy consumptions. In NoC-based (network on chip-based) multi-processor systems, distributing tasks to all processor elements could improve system utilization, but causing a large amount of inter-cluster communication and increasing data transmission energy consumption. Based on modeling the energy consumption model, this paper proposes an energy constraints task scheduling algorithm. Combining with the dependencies between tasks, the proposed algorithm dynamically calculates the cluster with minimal communication overhead and the cluster with maximal system utilization. Though calculating the energy consumption of corresponding clusters, the proposed algorithm chooses the plan which generates smaller energy consumption in order to balance system utilization and inter-cluster communication overhead. Simulation results show that the proposed algorithm can reduce the energy consumption, lower the inter-cluster communication overhead and increase the system utilization.