Aiming at the problem of resource allocation on energy-efficient in cognitive relay networks, a power allocation optimization algorithm based on the maximization of sub-user energy efficiency is proposed. A joint optimization model for spectrum sensing and transmission is established under multiple constraints. In the case of rate power consumption and interference limitation, the mixed integer nonlinear programming problem in NP problem is transformed into equivalent convex optimization problem by using fractional programming. The influence of different parameters on energy efficiency are analyzed. The simulation results show that the proposed power allocation optimization algorithm reduces the complexity of the algorithm while achieving the best energy efficiency.