This paper considers a non-orthogonal multiple access (NOMA) network consisting of a multi-antenna base station, multiple users and an information receiver. The working process of the network can be divided into two phases, namely downlink wireless power transfer phase and uplink information transmission phase. In the downlink wireless power transfer phase, the base station transfers power wirelessly to the users, and in the uplink information transmission phase, the users use their harvested energy to transmit information to the information receiver in a NOMA manner. When the users initial energy is low and thus cannot support its communication requirement, the wireless power transfer phase will be activated. We investigate the activation condition of the wireless power transfer phase. Furthermore, to maximize the throughput of uplink information transmission, we jointly optimize the time allocation between the downlink and uplink, the energy covariance matrix of wireless power transfer at the base station and transmit power of the users. Based on the Lagrangian multiplier method, we propose an optimal algorithm to find the solution to the considered problem in closed-form. Simulation results show that the proposed algorithm achieves much higher throughput performance than some benchmark schemes.