The growing deployment of Internet services in cloud data centers significantly increases the grid energy cost of cloud providers. Considering the environmental effect, many of current cloud providers migrate to green cloud data centers (GCDCs), and seek to reduce the usage of brown energy by partially (or entirely) adopting renewable energy sources. However, the temporal diversity in the grid price, wind speed and solar irradiance makes it a big challenge to minimize the grid energy cost of a GCDC while meeting the performance of each delay bounded request. This work proposes a Temporal Request Scheduling algorithm (TRS) that jointly considers the temporal diversity. TRS considers the long tail in real-life requests' delay, and can provide strict delay assurance to all arriving requests by scheduling them to execute within their delay bound. Besides, this work explicitly provides mathematical modeling of the relation between the service rate in a GCDC and the refusal of delay bounded requests. Specifically, TRS solves a constrained nonlinear optimization problem by a hybrid meta-heuristic in each of its iterations. Compared with some existing scheduling methods, TRS can achieve higher throughput and lower grid energy cost for a GCDC while meeting each request's delay requirement.