Recently, researchers pay much attention to Wireless Power Transfer-supported Device-to-Device (D2D) Communication underlaying Cellular Network (WPT-DCCN), which has reliable and controllable energy resource, high energy and spectral efficiency. However, the Energy Efficiency (EE) of the whole WPT-DCCN, which combines the time allocation and the power control of all users (including base station, cellular and D2D user), is rarely studied. Consequently, this study establishes an EE optimization model of the whole WPT-DCCN. Facing the complexity of the modeled mathematical model, the EE problem is approximately converted to a convex one by a proposed Convex Approximation Scheme (CAS). Due to the fractional expression of the variables needed to be solved, it is hard to directly obtain an optimal solution for the converted convex problem through current convex optimization toolbox such as CVX. Thus, we skillfully utilize the Generalized Benders Decomposition (GBD) to obtain an optimal solution for the converted convex problem by analyzing the characteristics of the converted problem. Simulation results show that the proposed CAS can give a tight lower-bound solution of the original EE problem. Furthermore, CAS can help us to investigate some meaningful system features, for example the suitable energy conversion efficiency setting under different WPT-DCCN scenarios, which can give us a clear reference about the deployment of future WPT-DCCN.