In this paper, we consider a single cell downlink non-orthogonal multiple access (NOMA) network and aim at maximizing the energy efficiency. The energy-efficient resource allocation problem is formulated as a non-convex and NP-hard problem. To decrease the computation complexity, we decouple the optimization problem as subchannel matching scheme and power allocation subproblems. We introduce a super-modular game and then design an algorithm to converge to the Nash equilibrium (NE) point. Then, a greedy subchannel matching algorithm with low complexity is given through a two-way choice between users and subchannels. However, for given subchannel matching scheme, power allocation is still a non- convex problem. We then transform the non-convex problem to a convex problem by applying a successive convex approximation method. Afterwards we provide an algorithm to converge to suboptimal solution by solving a convex problem iteratively. Finally, simulation result demonstrates that the energy efficiency performance of NOMA system is better than orthogonal frequency division multiple access (OFDMA) system.