The cost of grid energy of cloud providers rises greatly due to the growing number of Internet services. Therefore, researches on energy-efficient cloud data centers (ECDCs) are increasingly focused on the effective use of renewable energy instead of brown energy for environmental protection. Because of the temporal difference in price of grid, solar irradiance and wind speed, it is difficult to satisfy the performance of each delay bounded request with lower grid energy cost of an ECDC. In this work, a temporal request scheduling (TRS) algorithm is proposed for the temporal diversity with strict delay assurance to schedule the appropriate execution of all arriving requests. In addition, a mathematical model is presented to describe the relation between ECDC's service rate and request refusal. The formulated nonlinear optimization problem is solved by a hybrid meta-heuristic algorithm. Experimental results demonstrate that TRS can realize less cost of grid energy and higher throughput for an ECDC while satisfying requests' delay requirement than three representative scheduling methods.