The renewables will constitute an important part of the future smart grid. As a result, the growing portion of renewable generation in the power grid will bring challenges to the operations of the power grid because of the fluctuation and intermittency properties of renewables. In order to make the operations of power grid stable and reliable, the power outputs from renewable energy sources must be smoothed. In this paper, we propose a scheme inspired from the idea of the leaky bucket mechanism for smoothing the power output from a renewable energy system. In our proposed method, the settings of energy storage size and power output level have significant effects on the system performance and thus needs to be determined. An optimization framework is thus proposed for storage and power output planning of the renewable energy system. To operate our proposed scheme practically, a load-adaptive power smoothing algorithm is devised aiming to match the power output level with the actual load in the grid. Our simulation studies show that the proposed algorithm can reduce the operation cost comparing to other algorithms and maintain high renewable energy utilization.