In this letter, we investigate robust resource allocation for reconfigurable intelligent surface (RIS)-assisted wireless-powered communication networks, where the RIS assists energy transfer from a multi-antenna power station (PS) to multiple wireless-powered users (WPUs). WPUs communicate with an information receiver by an orthogonal frequency division multiple access mode. The total energy efficiency (EE) of WPUs is maximized by jointly optimizing the active beamforming of the PS, the subcarrier allocation factor, the energy harvesting time, the transmit power of WPUs, and the passive beamforming of the RIS under the constraints on the minimum harvested energy and the minimum throughput. To deal with the probabilistic constraints, Bernstein-type inequality is adopted. The transformed non-convex problem is converted into a convex one by using the Dinkelbach's method, matching theory, and the alternating optimization approach. Simulation results verify the effectiveness of the proposed algorithm in terms of robustness and EE.