The present work focuses on robust predictive control (RPC) of uncertain processes and proposes a new approach based on orthonormal series function modeling. In such unstructured modeling, the output signal is described as a weighted sum of orthonormal functions that uses approximative information about the time constant of the process. Due to an efficient uncertainty representation, this kind of modeling is advantageous in the RPC context, even for constrained systems and processes with integral action. The stability of the closed-loop system is guaranteed by the setting of sufficient conditions for the selection of the controller prediction horizon. Simulation results are presented to illustrate the performance of this new RPC algorithm.