This paper presents an feedback linearizing model predictive control scheme to design controllers for grid-connected systems with multiple photovoltaic (PV) units. The controllers are designed based on the generalized dynamical models of grid-connected systems with multiple PV units. The feedback linearization technique is applied to linearize the whole system which decouples the dynamical models of multiple PV units. The dynamic interactions among different PV units are considered as noises which are also decoupled through the proposed feedback linearization scheme. This decoupling allows to design linear controllers in a decentralized way for individual PV units. In this paper, a receding horizon-based model predictive scheme is employed to design linear controllers for feedback linearized PV units. The performance of the proposed scheme is evaluated on an IEEE test system under different operating conditions and compared with an existing PI controller in terms of maintaining unity power factor operation. Simulation results clearly show that the proposed controller performs in a much better way as compared to the PI controller.