In this paper, we study optimal power control for sequential NOMA in wireless relay networks. When a sequential NOMA scheme is used, in each time slot, the base station uses NOMA in the first phase, while the stronger relay uses NOMA in the second phase. To solve the problem of optimal power control for sequential NOMA, we design the geometric eight-point algorithm based on geometric properties of the non-convex optimization problem. Specifically, the proposed geometric eight-point algorithm adopts parabolas and the geometric classification algorithm to obtain a locally optimal power allocation vector with low computational complexity. Simulation results indicate that the proposed sequential NOMA scheme could increase the spectral efficiency and reduce the relay power consumption for wireless relay networks.